proxy: add /v1/responses and /v1/audio/voices endpoints (#448)

Updates #433
Fixes #442 #226

.coderabbit.yaml

@@ -8,8 +8,15 @@ reviews:
   poem: false
   review_status: true
   collapse_walkthrough: false
+  sequence_diagrams: false
+  finishing_touches:
+    docstrings:
+      enabled: false
   auto_review:
     enabled: true
     drafts: false
 chat:
   auto_reply: true
+issue_enrichment:
+  planning:
+    enabled: false

CLAUDE.md

@@ -11,8 +11,10 @@ llama-swap is a light weight, transparent proxy server that provides automatic m
 
 ## Testing
 
-- `make test-dev` - Use this when making iterative changes. Runs `go test` and `staticcheck`. Fix any static checking errors. Use this only when changes are made to any code under the `proxy/` directory
-- `make test-all` - runs at the end before completing work. Includes long running concurrency tests.
+- Follow test naming conventions like `TestProxyManager_<test name>`, `TestProcessGroup_<test name>`, etc.
+- Use `go test -v -run <name pattern for new tests>` to run any new tests you've written.
+- Use `make test-dev` after running new tests for a quick over all test run. This runs `go test` and `staticcheck`. Fix any static checking errors. Use this only when changes are made to any code under the `proxy/` directory
+- Use `make test-all` before completing work. This includes long running concurrency tests.
 
 ## Workflow Tasks
 

README.md

@@ -18,9 +18,13 @@ Built in Go for performance and simplicity, llama-swap has zero dependencies and
 - ✅ OpenAI API supported endpoints:
   - `v1/completions`
   - `v1/chat/completions`
+  - `v1/responses`
   - `v1/embeddings`
   - `v1/audio/speech` ([#36](https://github.com/mostlygeek/llama-swap/issues/36))
   - `v1/audio/transcriptions` ([docs](https://github.com/mostlygeek/llama-swap/issues/41#issuecomment-2722637867))
+  - `v1/audio/voices`
+  - `v1/images/generations`
+  - `v1/images/edits`
 - ✅ Anthropic API supported endpoints:
   - `v1/messages`
 - ✅ llama-server (llama.cpp) supported endpoints
@@ -34,6 +38,7 @@ Built in Go for performance and simplicity, llama-swap has zero dependencies and
   - `/running` - list currently running models ([#61](https://github.com/mostlygeek/llama-swap/issues/61))
   - `/log` - remote log monitoring
   - `/health` - just returns "OK"
+- ✅ API Key support - define keys to restrict access to API endpoints
 - ✅ Customizable
   - Run multiple models at once with `Groups` ([#107](https://github.com/mostlygeek/llama-swap/issues/107))
   - Automatic unloading of models after timeout by setting a `ttl`

config-schema.json

@@ -273,6 +273,58 @@
             },
             "additionalProperties": false,
             "description": "A dictionary of event triggers and actions. Only supported hook is on_startup."
+        },
+        "logToStdout": {
+            "type": "string",
+            "enum": [
+                "proxy",
+                "upstream",
+                "both",
+                "none"
+            ],
+            "default": "proxy",
+            "description": "Controls what is logged to stdout. 'proxy': logs generated by llama-swap, 'upstream': copy of upstream process stdout logs, 'both': both interleaved together, 'none': no logs written to stdout."
+        },
+        "apiKeys": {
+            "type": "array",
+            "items": {
+                "type": "string",
+                "minLength": 1
+            },
+            "default": [],
+            "description": "Require an API key when making requests to inference endpoints. When empty, authorization will not be checked. Each key is a non-empty string."
+        },
+        "peers": {
+            "type": "object",
+            "additionalProperties": {
+                "type": "object",
+                "required": [
+                    "proxy",
+                    "models"
+                ],
+                "properties": {
+                    "proxy": {
+                        "type": "string",
+                        "format": "uri",
+                        "description": "A valid base URL to proxy requests to. Requested path to llama-swap will be appended to the end of the proxy value."
+                    },
+                    "apiKey": {
+                        "type": "string",
+                        "default": "",
+                        "description": "A string key to be injected into the request. If blank, no key will be added. Key will be injected into headers: Authorization: Bearer <key> and x-api-key: <key>."
+                    },
+                    "models": {
+                        "type": "array",
+                        "items": {
+                            "type": "string",
+                            "minLength": 1
+                        },
+                        "description": "A list of models served by the peer."
+                    }
+                }
+            },
+            "default": {},
+            "description": "A dictionary of remote peers and models they provide. Peers can be another llama-swap or any server that provides the /v1/ generative API endpoints supported by llama-swap."
         }
     }
 }

config.example.yaml

@@ -34,6 +34,16 @@ logLevel: info
 # - For more info, read: https://pkg.go.dev/time#pkg-constants
 logTimeFormat: ""
 
+# logToStdout: controls what is logged to stdout
+# - optional, default: "proxy"
+# - valid values:
+#   - "proxy": logs generated by llama-swap when swapping models,
+#      handling requests, etc.
+#   - "upstream": a copy of an upstream processes stdout logs
+#   - "both": both the proxy and upstream logs interleaved together
+#   - "none": no logs are ever written to stdout
+logToStdout: "proxy"
+
 # metricsMaxInMemory: maximum number of metrics to keep in memory
 # - optional, default: 1000
 # - controls how many metrics are stored in memory before older ones are discarded
@@ -60,6 +70,16 @@ sendLoadingState: true
 #   all fields except for Id so chat UIs can use the alias equivalent to the original.
 includeAliasesInList: false
 
+# apiKeys: require an API key when making requests to inference endpoints
+# - optional, default: []
+# - when empty (the default) authorization will not be checked as llama-swap is default-allow
+# - each key is a non-empty string
+apiKeys:
+  - "sk-hunter2"
+  # hint, one liner: printf "sk-%s\n" "$(head -c 48 /dev/urandom | base64 )"
+  - "sk-gyCPiKUcIfPlaM4OSMZekkprgijPx6+OsmQs8Rsg0xZ9qpy6gKWsIKqHOk+cgXVx"
+  - "sk-+QtIn0Zjj4UHjiaZYiZEnru4mrwKM9RzhmJeK5SobNXLl8QMFXxGz1/2lEuvQpkb"
+
 # macros: a dictionary of string substitutions
 # - optional, default: empty dictionary
 # - macros are reusable snippets
@@ -321,3 +341,35 @@ hooks:
     #   otherwise models will be loaded and swapped out
     preload:
       - "llama"
+
+# peers: a dictionary of remote peers and models they provide
+# - optional, default empty dictionary
+# - peers can be another llama-swap
+# - peers can be any server that provides the /v1/ generative api endpoints supported by llama-swap
+peers:
+  # keys is the peer'd ID
+  llama-swap-peer:
+    # proxy: a valid base URL to proxy requests to
+    # - required
+    # - requested path to llama-swap will be appended to the end of the proxy value
+    proxy: http://192.168.1.23
+    # models: a list of models served by the peer
+    # - required
+    models:
+      - model_a
+      - model_b
+      - embeddings/model_c
+  openrouter:
+    proxy: https://openrouter.ai/api
+    # apiKey: a string key to be injected into the request
+    # - optional, default: ""
+    # - if blank, no key will be added to the request
+    # - key will be injected into headers: Authorization: Bearer <key> and x-api-key: <key>
+    apiKey: sk-your-openrouter-key
+    models:
+      - meta-llama/llama-3.1-8b-instruct
+      - qwen/qwen3-235b-a22b-2507
+      - deepseek/deepseek-v3.2
+      - z-ai/glm-4.7
+      - moonshotai/kimi-k2-0905
+      - minimax/minimax-m2.1

docs/configuration.md

@@ -86,9 +86,12 @@ llama-swap supports many more features to customize how you want to manage your
 ## Full Configuration Example
 
 > [!NOTE]
-> This is a copy of `config.example.yaml`. Always check that for the most up to date examples.
+> Always check [config.example.yaml](https://github.com/mostlygeek/llama-swap/blob/main/config.example.yaml) for the most up to date reference for all example configurations.
 
 ```yaml
+# add this modeline for validation in vscode
+# yaml-language-server: $schema=https://raw.githubusercontent.com/mostlygeek/llama-swap/refs/heads/main/config-schema.json
+#
 # llama-swap YAML configuration example
 # -------------------------------------
 #
@@ -114,6 +117,24 @@ healthCheckTimeout: 500
 # - Valid log levels: debug, info, warn, error
 logLevel: info
 
+# logTimeFormat: enables and sets the logging timestamp format
+# - optional, default (disabled): ""
+# - Valid values: "", "ansic", "unixdate", "rubydate", "rfc822", "rfc822z",
+#   "rfc850", "rfc1123", "rfc1123z", "rfc3339", "rfc3339nano", "kitchen",
+#   "stamp", "stampmilli", "stampmicro", and "stampnano".
+# - For more info, read: https://pkg.go.dev/time#pkg-constants
+logTimeFormat: ""
+
+# logToStdout: controls what is logged to stdout
+# - optional, default: "proxy"
+# - valid values:
+#   - "proxy": logs generated by llama-swap when swapping models,
+#      handling requests, etc.
+#   - "upstream": a copy of an upstream processes stdout logs
+#   - "both": both the proxy and upstream logs interleaved together
+#   - "none": no logs are ever written to stdout
+logToStdout: "proxy"
+
 # metricsMaxInMemory: maximum number of metrics to keep in memory
 # - optional, default: 1000
 # - controls how many metrics are stored in memory before older ones are discarded
@@ -126,6 +147,30 @@ metricsMaxInMemory: 1000
 # - it is automatically incremented for every model that uses it
 startPort: 10001
 
+# sendLoadingState: inject loading status updates into the reasoning (thinking)
+# field
+# - optional, default: false
+# - when true, a stream of loading messages will be sent to the client in the
+#   reasoning field so chat UIs can show that loading is in progress.
+# - see #366 for more details
+sendLoadingState: true
+
+# includeAliasesInList: present aliases within the /v1/models OpenAI API listing
+# - optional, default: false
+# - when true, model aliases will be output to the API model listing duplicating
+#   all fields except for Id so chat UIs can use the alias equivalent to the original.
+includeAliasesInList: false
+
+# apiKeys: require an API key when making requests to inference endpoints
+# - optional, default: []
+# - when empty (the default) authorization will not be checked as llama-swap is default-allow
+# - each key is a non-empty string
+apiKeys:
+  - "sk-hunter2"
+  # hint, one liner: printf "sk-%s\n" "$(head -c 48 /dev/urandom | base64 )"
+  - "sk-gyCPiKUcIfPlaM4OSMZekkprgijPx6+OsmQs8Rsg0xZ9qpy6gKWsIKqHOk+cgXVx"
+  - "sk-+QtIn0Zjj4UHjiaZYiZEnru4mrwKM9RzhmJeK5SobNXLl8QMFXxGz1/2lEuvQpkb"
+
 # macros: a dictionary of string substitutions
 # - optional, default: empty dictionary
 # - macros are reusable snippets
@@ -274,6 +319,10 @@ models:
     # - recommended to be omitted and the default used
     concurrencyLimit: 0
 
+    # sendLoadingState: overrides the global sendLoadingState setting for this model
+    # - optional, default: undefined (use global setting)
+    sendLoadingState: false
+
   # Unlisted model example:
   "qwen-unlisted":
     # unlisted: boolean, true or false
@@ -383,4 +432,36 @@ hooks:
     #   otherwise models will be loaded and swapped out
     preload:
       - "llama"
+
+# peers: a dictionary of remote peers and models they provide
+# - optional, default empty dictionary
+# - peers can be another llama-swap
+# - peers can be any server that provides the /v1/ generative api endpoints supported by llama-swap
+peers:
+  # keys is the peer'd ID
+  llama-swap-peer:
+    # proxy: a valid base URL to proxy requests to
+    # - required
+    # - requested path to llama-swap will be appended to the end of the proxy value
+    proxy: http://192.168.1.23
+    # models: a list of models served by the peer
+    # - required
+    models:
+      - model_a
+      - model_b
+      - embeddings/model_c
+  openrouter:
+    proxy: https://openrouter.ai/api
+    # apiKey: a string key to be injected into the request
+    # - optional, default: ""
+    # - if blank, no key will be added to the request
+    # - key will be injected into headers: Authorization: Bearer <key> and x-api-key: <key>
+    apiKey: sk-your-openrouter-key
+    models:
+      - meta-llama/llama-3.1-8b-instruct
+      - qwen/qwen3-235b-a22b-2507
+      - deepseek/deepseek-v3.2
+      - z-ai/glm-4.7
+      - moonshotai/kimi-k2-0905
+      - minimax/minimax-m2.1
 ```

proxy/config/config.go

@@ -15,6 +15,12 @@ import (
 )
 
 const DEFAULT_GROUP_ID = "(default)"
+const (
+	LogToStdoutProxy    = "proxy"
+	LogToStdoutUpstream = "upstream"
+	LogToStdoutBoth     = "both"
+	LogToStdoutNone     = "none"
+)
 
 type MacroEntry struct {
 	Name  string
@@ -114,6 +120,7 @@ type Config struct {
 	LogRequests        bool                   `yaml:"logRequests"`
 	LogLevel           string                 `yaml:"logLevel"`
 	LogTimeFormat      string                 `yaml:"logTimeFormat"`
+	LogToStdout        string                 `yaml:"logToStdout"`
 	MetricsMaxInMemory int                    `yaml:"metricsMaxInMemory"`
 	Models             map[string]ModelConfig `yaml:"models"` /* key is model ID */
 	Profiles           map[string][]string    `yaml:"profiles"`
@@ -136,6 +143,12 @@ type Config struct {
 
 	// present aliases to /v1/models OpenAI API listing
 	IncludeAliasesInList bool `yaml:"includeAliasesInList"`
+
+	// support API keys, see issue #433, #50, #251
+	RequiredAPIKeys []string `yaml:"apiKeys"`
+
+	// support remote peers, see issue #433, #296
+	Peers PeerDictionaryConfig `yaml:"peers"`
 }
 
 func (c *Config) RealModelName(search string) (string, bool) {
@@ -177,6 +190,7 @@ func LoadConfigFromReader(r io.Reader) (Config, error) {
 		StartPort:          5800,
 		LogLevel:           "info",
 		LogTimeFormat:      "",
+		LogToStdout:        LogToStdoutProxy,
 		MetricsMaxInMemory: 1000,
 	}
 	err = yaml.Unmarshal(data, &config)
@@ -193,6 +207,12 @@ func LoadConfigFromReader(r io.Reader) (Config, error) {
 		return Config{}, fmt.Errorf("startPort must be greater than 1")
 	}
 
+	switch config.LogToStdout {
+	case LogToStdoutProxy, LogToStdoutUpstream, LogToStdoutBoth, LogToStdoutNone:
+	default:
+		return Config{}, fmt.Errorf("logToStdout must be one of: proxy, upstream, both, none")
+	}
+
 	// Populate the aliases map
 	config.aliases = make(map[string]string)
 	for modelName, modelConfig := range config.Models {
@@ -404,6 +424,17 @@ func LoadConfigFromReader(r io.Reader) (Config, error) {
 		config.Hooks.OnStartup.Preload = toPreload
 	}
 
+	// check api keys validatity
+	for _, apikey := range config.RequiredAPIKeys {
+		if apikey == "" {
+			return Config{}, fmt.Errorf("empty api key found in apiKeys")
+		}
+
+		if strings.Contains(apikey, " ") {
+			return Config{}, fmt.Errorf("api key cannot contain spaces: `%s`", apikey)
+		}
+	}
+
 	return config, nil
 }
 

proxy/config/config_posix_test.go

@@ -166,6 +166,7 @@ groups:
 	expected := Config{
 		LogLevel:      "info",
 		LogTimeFormat: "",
+		LogToStdout:   LogToStdoutProxy,
 		StartPort:     5800,
 		Macros: MacroList{
 			{"svr-path", "path/to/server"},

proxy/config/config_test.go

@@ -761,3 +761,51 @@ models:
 		})
 	}
 }
+
+func TestConfig_APIKeys_Invalid(t *testing.T) {
+	tests := []struct {
+		name        string
+		content     string
+		expectedErr string
+	}{
+		{
+			name:        "empty string",
+			content:     `apiKeys: [""]`,
+			expectedErr: "empty api key found in apiKeys",
+		},
+		{
+			name:        "blank spaces only",
+			content:     `apiKeys: ["   "]`,
+			expectedErr: "api key cannot contain spaces: `   `",
+		},
+		{
+			name:        "contains leading space",
+			content:     `apiKeys: [" key123"]`,
+			expectedErr: "api key cannot contain spaces: ` key123`",
+		},
+		{
+			name:        "contains trailing space",
+			content:     `apiKeys: ["key123 "]`,
+			expectedErr: "api key cannot contain spaces: `key123 `",
+		},
+		{
+			name:        "contains middle space",
+			content:     `apiKeys: ["key 123"]`,
+			expectedErr: "api key cannot contain spaces: `key 123`",
+		},
+		{
+			name:        "empty in list with valid keys",
+			content:     `apiKeys: ["valid-key", "", "another-key"]`,
+			expectedErr: "empty api key found in apiKeys",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := LoadConfigFromReader(strings.NewReader(tt.content))
+			if assert.Error(t, err) {
+				assert.Equal(t, tt.expectedErr, err.Error())
+			}
+		})
+	}
+}

proxy/config/config_windows_test.go

@@ -158,6 +158,7 @@ groups:
 	expected := Config{
 		LogLevel:      "info",
 		LogTimeFormat: "",
+		LogToStdout:   LogToStdoutProxy,
 		StartPort:     5800,
 		Macros: MacroList{
 			{"svr-path", "path/to/server"},

proxy/config/peer.go

@@ -0,0 +1,47 @@
+package config
+
+import (
+	"fmt"
+	"net/url"
+)
+
+type PeerDictionaryConfig map[string]PeerConfig
+type PeerConfig struct {
+	Proxy    string   `yaml:"proxy"`
+	ProxyURL *url.URL `yaml:"-"`
+	ApiKey   string   `yaml:"apiKey"`
+	Models   []string `yaml:"models"`
+}
+
+func (c *PeerConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawPeerConfig PeerConfig
+	defaults := rawPeerConfig{
+		Proxy:  "",
+		ApiKey: "",
+		Models: []string{},
+	}
+
+	if err := unmarshal(&defaults); err != nil {
+		return err
+	}
+
+	// Validate proxy is not empty
+	if defaults.Proxy == "" {
+		return fmt.Errorf("proxy is required")
+	}
+
+	// Validate proxy is a valid URL and store the parsed value
+	parsedURL, err := url.Parse(defaults.Proxy)
+	if err != nil {
+		return fmt.Errorf("invalid peer proxy URL (%s): %w", defaults.Proxy, err)
+	}
+	defaults.ProxyURL = parsedURL
+
+	// Validate models is not empty
+	if len(defaults.Models) == 0 {
+		return fmt.Errorf("peer models can not be empty")
+	}
+
+	*c = PeerConfig(defaults)
+	return nil
+}

proxy/config/peer_test.go

@@ -0,0 +1,139 @@
+package config
+
+import (
+	"testing"
+
+	"gopkg.in/yaml.v3"
+)
+
+func TestPeerConfig_UnmarshalYAML(t *testing.T) {
+	tests := []struct {
+		name    string
+		yaml    string
+		wantErr string
+	}{
+		{
+			name: "valid config",
+			yaml: `
+proxy: http://192.168.1.23
+models:
+  - model_a
+  - model_b
+`,
+			wantErr: "",
+		},
+		{
+			name: "valid config with apiKey",
+			yaml: `
+proxy: https://openrouter.ai/api
+apiKey: sk-test-key
+models:
+  - meta-llama/llama-3.1-8b-instruct
+`,
+			wantErr: "",
+		},
+		{
+			name: "missing proxy",
+			yaml: `
+models:
+  - model_a
+`,
+			wantErr: "proxy is required",
+		},
+		{
+			name: "empty proxy",
+			yaml: `
+proxy: ""
+models:
+  - model_a
+`,
+			wantErr: "proxy is required",
+		},
+		{
+			name: "invalid proxy URL",
+			yaml: `
+proxy: "://invalid"
+models:
+  - model_a
+`,
+			wantErr: "invalid peer proxy URL",
+		},
+		{
+			name: "missing models",
+			yaml: `
+proxy: http://localhost:8080
+`,
+			wantErr: "peer models can not be empty",
+		},
+		{
+			name: "empty models",
+			yaml: `
+proxy: http://localhost:8080
+models: []
+`,
+			wantErr: "peer models can not be empty",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			var config PeerConfig
+			err := yaml.Unmarshal([]byte(tt.yaml), &config)
+
+			if tt.wantErr == "" {
+				if err != nil {
+					t.Errorf("unexpected error: %v", err)
+				}
+			} else {
+				if err == nil {
+					t.Errorf("expected error containing %q, got nil", tt.wantErr)
+				} else if !contains(err.Error(), tt.wantErr) {
+					t.Errorf("expected error containing %q, got %q", tt.wantErr, err.Error())
+				}
+			}
+		})
+	}
+}
+
+func TestPeerConfig_ProxyURL(t *testing.T) {
+	yamlData := `
+proxy: http://192.168.1.23:8080/api
+apiKey: sk-test
+models:
+  - model_a
+`
+	var config PeerConfig
+	err := yaml.Unmarshal([]byte(yamlData), &config)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	if config.ProxyURL == nil {
+		t.Fatal("ProxyURL should not be nil")
+	}
+
+	if config.ProxyURL.Host != "192.168.1.23:8080" {
+		t.Errorf("expected host %q, got %q", "192.168.1.23:8080", config.ProxyURL.Host)
+	}
+
+	if config.ProxyURL.Scheme != "http" {
+		t.Errorf("expected scheme %q, got %q", "http", config.ProxyURL.Scheme)
+	}
+
+	if config.ProxyURL.Path != "/api" {
+		t.Errorf("expected path %q, got %q", "/api", config.ProxyURL.Path)
+	}
+}
+
+func contains(s, substr string) bool {
+	return len(s) >= len(substr) && searchSubstring(s, substr)
+}
+
+func searchSubstring(s, substr string) bool {
+	for i := 0; i <= len(s)-len(substr); i++ {
+		if s[i:i+len(substr)] == substr {
+			return true
+		}
+	}
+	return false
+}

proxy/metrics_monitor.go

@@ -2,6 +2,8 @@ package proxy
 
 import (
 	"bytes"
+	"compress/flate"
+	"compress/gzip"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -96,6 +98,12 @@ func (mp *metricsMonitor) wrapHandler(
 	next func(modelID string, w http.ResponseWriter, r *http.Request) error,
 ) error {
 	recorder := newBodyCopier(writer)
+
+	// Filter Accept-Encoding to only include encodings we can decompress for metrics
+	if ae := request.Header.Get("Accept-Encoding"); ae != "" {
+		request.Header.Set("Accept-Encoding", filterAcceptEncoding(ae))
+	}
+
 	if err := next(modelID, recorder, request); err != nil {
 		return err
 	}
@@ -108,17 +116,36 @@ func (mp *metricsMonitor) wrapHandler(
 		return nil
 	}
 
+	// Initialize default metrics - these will always be recorded
+	tm := TokenMetrics{
+		Timestamp:  time.Now(),
+		Model:      modelID,
+		DurationMs: int(time.Since(recorder.StartTime()).Milliseconds()),
+	}
+
 	body := recorder.body.Bytes()
 	if len(body) == 0 {
-		mp.logger.Warn("metrics skipped, empty body")
+		mp.logger.Warn("metrics: empty body, recording minimal metrics")
+		mp.addMetrics(tm)
 		return nil
 	}
 
-	if strings.Contains(recorder.Header().Get("Content-Type"), "text/event-stream") {
-		if tm, err := processStreamingResponse(modelID, recorder.StartTime(), body); err != nil {
-			mp.logger.Warnf("error processing streaming response: %v, path=%s", err, request.URL.Path)
-		} else {
+	// Decompress if needed
+	if encoding := recorder.Header().Get("Content-Encoding"); encoding != "" {
+		var err error
+		body, err = decompressBody(body, encoding)
+		if err != nil {
+			mp.logger.Warnf("metrics: decompression failed: %v, path=%s, recording minimal metrics", err, request.URL.Path)
 			mp.addMetrics(tm)
+			return nil
+		}
+	}
+
+	if strings.Contains(recorder.Header().Get("Content-Type"), "text/event-stream") {
+		if parsed, err := processStreamingResponse(modelID, recorder.StartTime(), body); err != nil {
+			mp.logger.Warnf("error processing streaming response: %v, path=%s, recording minimal metrics", err, request.URL.Path)
+		} else {
+			tm = parsed
 		}
 	} else {
 		if gjson.ValidBytes(body) {
@@ -127,18 +154,18 @@ func (mp *metricsMonitor) wrapHandler(
 			timings := parsed.Get("timings")
 
 			if usage.Exists() || timings.Exists() {
-				if tm, err := parseMetrics(modelID, recorder.StartTime(), usage, timings); err != nil {
-					mp.logger.Warnf("error parsing metrics: %v, path=%s", err, request.URL.Path)
+				if parsedMetrics, err := parseMetrics(modelID, recorder.StartTime(), usage, timings); err != nil {
+					mp.logger.Warnf("error parsing metrics: %v, path=%s, recording minimal metrics", err, request.URL.Path)
 				} else {
-					mp.addMetrics(tm)
+					tm = parsedMetrics
 				}
 			}
-
 		} else {
-			mp.logger.Warnf("metrics skipped, invalid JSON in response body path=%s", request.URL.Path)
+			mp.logger.Warnf("metrics: invalid JSON in response body path=%s, recording minimal metrics", request.URL.Path)
 		}
 	}
 
+	mp.addMetrics(tm)
 	return nil
 }
 
@@ -251,6 +278,25 @@ func parseMetrics(modelID string, start time.Time, usage, timings gjson.Result)
 	}, nil
 }
 
+// decompressBody decompresses the body based on Content-Encoding header
+func decompressBody(body []byte, encoding string) ([]byte, error) {
+	switch strings.ToLower(strings.TrimSpace(encoding)) {
+	case "gzip":
+		reader, err := gzip.NewReader(bytes.NewReader(body))
+		if err != nil {
+			return nil, err
+		}
+		defer reader.Close()
+		return io.ReadAll(reader)
+	case "deflate":
+		reader := flate.NewReader(bytes.NewReader(body))
+		defer reader.Close()
+		return io.ReadAll(reader)
+	default:
+		return body, nil // Return as-is for unknown/no encoding
+	}
+}
+
 // responseBodyCopier records the response body and writes to the original response writer
 // while also capturing it in a buffer for later processing
 type responseBodyCopier struct {
@@ -289,3 +335,25 @@ func (w *responseBodyCopier) Header() http.Header {
 func (w *responseBodyCopier) StartTime() time.Time {
 	return w.start
 }
+
+// filterAcceptEncoding filters the Accept-Encoding header to only include
+// encodings we can decompress (gzip, deflate). This respects the client's
+// preferences while ensuring we can parse response bodies for metrics.
+func filterAcceptEncoding(acceptEncoding string) string {
+	if acceptEncoding == "" {
+		return ""
+	}
+
+	supported := map[string]bool{"gzip": true, "deflate": true}
+	var filtered []string
+
+	for _, part := range strings.Split(acceptEncoding, ",") {
+		// Parse encoding and optional quality value (e.g., "gzip;q=1.0")
+		encoding := strings.TrimSpace(strings.Split(part, ";")[0])
+		if supported[strings.ToLower(encoding)] {
+			filtered = append(filtered, strings.TrimSpace(part))
+		}
+	}
+
+	return strings.Join(filtered, ", ")
+}

proxy/metrics_monitor_test.go

@@ -1,6 +1,9 @@
 package proxy
 
 import (
+	"bytes"
+	"compress/flate"
+	"compress/gzip"
 	"encoding/json"
 	"net/http"
 	"net/http/httptest"
@@ -291,7 +294,7 @@ data: [DONE]
 		assert.Equal(t, 0, len(metrics))
 	})
 
-	t.Run("empty response body does not record metrics", func(t *testing.T) {
+	t.Run("empty response body records minimal metrics", func(t *testing.T) {
 		mm := newMetricsMonitor(testLogger, 10)
 
 		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
@@ -307,10 +310,13 @@ data: [DONE]
 		assert.NoError(t, err)
 
 		metrics := mm.getMetrics()
-		assert.Equal(t, 0, len(metrics))
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
 	})
 
-	t.Run("invalid JSON does not record metrics", func(t *testing.T) {
+	t.Run("invalid JSON records minimal metrics", func(t *testing.T) {
 		mm := newMetricsMonitor(testLogger, 10)
 
 		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
@@ -328,7 +334,10 @@ data: [DONE]
 		assert.NoError(t, err) // Errors after response is sent are logged, not returned
 
 		metrics := mm.getMetrics()
-		assert.Equal(t, 0, len(metrics))
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
 	})
 
 	t.Run("next handler error is propagated", func(t *testing.T) {
@@ -350,7 +359,7 @@ data: [DONE]
 		assert.Equal(t, 0, len(metrics))
 	})
 
-	t.Run("response without usage or timings does not record metrics", func(t *testing.T) {
+	t.Run("response without usage or timings records minimal metrics", func(t *testing.T) {
 		mm := newMetricsMonitor(testLogger, 10)
 
 		responseBody := `{"result": "ok"}`
@@ -367,10 +376,13 @@ data: [DONE]
 		ginCtx, _ := gin.CreateTestContext(rec)
 
 		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
-		assert.NoError(t, err) // Errors after response is sent are logged, not returned
+		assert.NoError(t, err)
 
 		metrics := mm.getMetrics()
-		assert.Equal(t, 0, len(metrics))
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
 	})
 }
 
@@ -598,7 +610,7 @@ data: [DONE]
 		assert.Equal(t, 50, metrics[0].OutputTokens)
 	})
 
-	t.Run("handles streaming with no valid JSON", func(t *testing.T) {
+	t.Run("handles streaming with no valid JSON records minimal metrics", func(t *testing.T) {
 		mm := newMetricsMonitor(testLogger, 10)
 
 		responseBody := `data: not json
@@ -619,13 +631,16 @@ data: [DONE]
 		ginCtx, _ := gin.CreateTestContext(rec)
 
 		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
-		assert.NoError(t, err) // Errors after response is sent are logged, not returned
+		assert.NoError(t, err)
 
 		metrics := mm.getMetrics()
-		assert.Equal(t, 0, len(metrics))
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
 	})
 
-	t.Run("handles empty streaming response", func(t *testing.T) {
+	t.Run("handles empty streaming response records minimal metrics", func(t *testing.T) {
 		mm := newMetricsMonitor(testLogger, 10)
 
 		responseBody := ``
@@ -642,11 +657,13 @@ data: [DONE]
 		ginCtx, _ := gin.CreateTestContext(rec)
 
 		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
-		// Empty body should not trigger WrapHandler processing
 		assert.NoError(t, err)
 
 		metrics := mm.getMetrics()
-		assert.Equal(t, 0, len(metrics))
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
 	})
 }
 
@@ -691,3 +708,127 @@ func BenchmarkMetricsMonitor_AddMetrics_SmallBuffer(b *testing.B) {
 		mm.addMetrics(metric)
 	}
 }
+
+func TestMetricsMonitor_WrapHandler_Compression(t *testing.T) {
+	t.Run("gzip encoded response", func(t *testing.T) {
+		mm := newMetricsMonitor(testLogger, 10)
+
+		responseBody := `{"usage": {"prompt_tokens": 100, "completion_tokens": 50}}`
+
+		// Compress with gzip
+		var buf bytes.Buffer
+		gzWriter := gzip.NewWriter(&buf)
+		gzWriter.Write([]byte(responseBody))
+		gzWriter.Close()
+		compressedBody := buf.Bytes()
+
+		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
+			w.Header().Set("Content-Type", "application/json")
+			w.Header().Set("Content-Encoding", "gzip")
+			w.WriteHeader(http.StatusOK)
+			w.Write(compressedBody)
+			return nil
+		}
+
+		req := httptest.NewRequest("POST", "/test", nil)
+		rec := httptest.NewRecorder()
+		ginCtx, _ := gin.CreateTestContext(rec)
+
+		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
+		assert.NoError(t, err)
+
+		metrics := mm.getMetrics()
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 100, metrics[0].InputTokens)
+		assert.Equal(t, 50, metrics[0].OutputTokens)
+	})
+
+	t.Run("deflate encoded response", func(t *testing.T) {
+		mm := newMetricsMonitor(testLogger, 10)
+
+		responseBody := `{"usage": {"prompt_tokens": 200, "completion_tokens": 75}}`
+
+		// Compress with deflate
+		var buf bytes.Buffer
+		flateWriter, _ := flate.NewWriter(&buf, flate.DefaultCompression)
+		flateWriter.Write([]byte(responseBody))
+		flateWriter.Close()
+		compressedBody := buf.Bytes()
+
+		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
+			w.Header().Set("Content-Type", "application/json")
+			w.Header().Set("Content-Encoding", "deflate")
+			w.WriteHeader(http.StatusOK)
+			w.Write(compressedBody)
+			return nil
+		}
+
+		req := httptest.NewRequest("POST", "/test", nil)
+		rec := httptest.NewRecorder()
+		ginCtx, _ := gin.CreateTestContext(rec)
+
+		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
+		assert.NoError(t, err)
+
+		metrics := mm.getMetrics()
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 200, metrics[0].InputTokens)
+		assert.Equal(t, 75, metrics[0].OutputTokens)
+	})
+
+	t.Run("invalid gzip data records minimal metrics", func(t *testing.T) {
+		mm := newMetricsMonitor(testLogger, 10)
+
+		// Invalid compressed data
+		invalidData := []byte("this is not gzip data")
+
+		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
+			w.Header().Set("Content-Type", "application/json")
+			w.Header().Set("Content-Encoding", "gzip")
+			w.WriteHeader(http.StatusOK)
+			w.Write(invalidData)
+			return nil
+		}
+
+		req := httptest.NewRequest("POST", "/test", nil)
+		rec := httptest.NewRecorder()
+		ginCtx, _ := gin.CreateTestContext(rec)
+
+		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
+		assert.NoError(t, err) // Should not return error, just log warning
+
+		metrics := mm.getMetrics()
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, "test-model", metrics[0].Model)
+		assert.Equal(t, 0, metrics[0].InputTokens)
+		assert.Equal(t, 0, metrics[0].OutputTokens)
+	})
+
+	t.Run("unknown encoding treated as uncompressed", func(t *testing.T) {
+		mm := newMetricsMonitor(testLogger, 10)
+
+		responseBody := `{"usage": {"prompt_tokens": 300, "completion_tokens": 100}}`
+
+		nextHandler := func(modelID string, w http.ResponseWriter, r *http.Request) error {
+			w.Header().Set("Content-Type", "application/json")
+			w.Header().Set("Content-Encoding", "unknown-encoding")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte(responseBody))
+			return nil
+		}
+
+		req := httptest.NewRequest("POST", "/test", nil)
+		rec := httptest.NewRecorder()
+		ginCtx, _ := gin.CreateTestContext(rec)
+
+		err := mm.wrapHandler("test-model", ginCtx.Writer, req, nextHandler)
+		assert.NoError(t, err)
+
+		metrics := mm.getMetrics()
+		assert.Equal(t, 1, len(metrics))
+		assert.Equal(t, 300, metrics[0].InputTokens)
+		assert.Equal(t, 100, metrics[0].OutputTokens)
+	})
+}

proxy/peerproxy.go

@@ -0,0 +1,127 @@
+package proxy
+
+import (
+	"fmt"
+	"net"
+	"net/http"
+	"net/http/httputil"
+	"runtime"
+	"sort"
+	"strings"
+	"time"
+
+	"github.com/mostlygeek/llama-swap/proxy/config"
+)
+
+type peerProxyMember struct {
+	peerID       string
+	reverseProxy *httputil.ReverseProxy
+	apiKey       string
+}
+
+type PeerProxy struct {
+	peers    config.PeerDictionaryConfig
+	proxyMap map[string]*peerProxyMember
+}
+
+func NewPeerProxy(peers config.PeerDictionaryConfig, proxyLogger *LogMonitor) (*PeerProxy, error) {
+	proxyMap := make(map[string]*peerProxyMember)
+
+	// Sort peer IDs for consistent iteration order
+	peerIDs := make([]string, 0, len(peers))
+	for peerID := range peers {
+		peerIDs = append(peerIDs, peerID)
+	}
+	sort.Strings(peerIDs)
+
+	// Create a shared transport with reasonable timeouts for peer connections
+	// these can be tuned with feedback later
+	peerTransport := &http.Transport{
+		DialContext: (&net.Dialer{
+			Timeout:   30 * time.Second, // Connection timeout
+			KeepAlive: 30 * time.Second,
+		}).DialContext,
+		TLSHandshakeTimeout:   10 * time.Second,
+		ResponseHeaderTimeout: 60 * time.Second, // Time to wait for response headers
+		ExpectContinueTimeout: 1 * time.Second,
+		MaxIdleConns:          100,
+		MaxIdleConnsPerHost:   10,
+		IdleConnTimeout:       90 * time.Second,
+	}
+
+	for _, peerID := range peerIDs {
+		peer := peers[peerID]
+		// Create reverse proxy for this peer
+		reverseProxy := httputil.NewSingleHostReverseProxy(peer.ProxyURL)
+		reverseProxy.Transport = peerTransport
+
+		// Wrap Director to set Host header for remote hosts (not localhost)
+		originalDirector := reverseProxy.Director
+		reverseProxy.Director = func(req *http.Request) {
+			originalDirector(req)
+			// Ensure Host header matches target URL for remote proxying
+			req.Host = req.URL.Host
+		}
+
+		reverseProxy.ModifyResponse = func(resp *http.Response) error {
+			if strings.Contains(strings.ToLower(resp.Header.Get("Content-Type")), "text/event-stream") {
+				resp.Header.Set("X-Accel-Buffering", "no")
+			}
+			return nil
+		}
+
+		reverseProxy.ErrorHandler = func(w http.ResponseWriter, r *http.Request, err error) {
+			proxyLogger.Warnf("peer %s: proxy error: %v", peerID, err)
+			errMsg := fmt.Sprintf("peer proxy error: %v", err)
+			if runtime.GOOS == "darwin" && strings.Contains(err.Error(), "connect: no route to host") {
+				errMsg += " (hint: on macOS, check System Settings > Privacy & Security > Local Network permissions)"
+			}
+			http.Error(w, errMsg, http.StatusBadGateway)
+		}
+
+		pp := &peerProxyMember{
+			peerID:       peerID,
+			reverseProxy: reverseProxy,
+			apiKey:       peer.ApiKey,
+		}
+
+		// Map each model to this peer's proxy
+		for _, modelID := range peer.Models {
+			if _, found := proxyMap[modelID]; found {
+				proxyLogger.Warnf("peer %s: model %s already mapped to another peer, skipping", peerID, modelID)
+				continue
+			}
+			proxyMap[modelID] = pp
+		}
+	}
+
+	return &PeerProxy{
+		peers:    peers,
+		proxyMap: proxyMap,
+	}, nil
+}
+
+func (p *PeerProxy) HasPeerModel(modelID string) bool {
+	_, found := p.proxyMap[modelID]
+	return found
+}
+
+func (p *PeerProxy) ListPeers() config.PeerDictionaryConfig {
+	return p.peers
+}
+
+func (p *PeerProxy) ProxyRequest(model_id string, writer http.ResponseWriter, request *http.Request) error {
+	pp, found := p.proxyMap[model_id]
+	if !found {
+		return fmt.Errorf("no peer proxy found for model %s", model_id)
+	}
+
+	// Inject API key if configured for this peer
+	if pp.apiKey != "" {
+		request.Header.Set("Authorization", "Bearer "+pp.apiKey)
+		request.Header.Set("x-api-key", pp.apiKey)
+	}
+
+	pp.reverseProxy.ServeHTTP(writer, request)
+	return nil
+}

proxy/peerproxy_test.go

@@ -0,0 +1,268 @@
+package proxy
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"strings"
+	"testing"
+
+	"github.com/mostlygeek/llama-swap/proxy/config"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestNewPeerProxy_EmptyPeers(t *testing.T) {
+	peers := config.PeerDictionaryConfig{}
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+	assert.NotNil(t, pm)
+	assert.Empty(t, pm.proxyMap)
+}
+
+func TestNewPeerProxy_SinglePeer(t *testing.T) {
+	proxyURL, _ := url.Parse("http://peer1.example.com:8080")
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    "http://peer1.example.com:8080",
+			ProxyURL: proxyURL,
+			ApiKey:   "test-key",
+			Models:   []string{"model-a", "model-b"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+	assert.Len(t, pm.proxyMap, 2)
+	assert.True(t, pm.HasPeerModel("model-a"))
+	assert.True(t, pm.HasPeerModel("model-b"))
+	assert.False(t, pm.HasPeerModel("model-c"))
+}
+
+func TestNewPeerProxy_MultiplePeers(t *testing.T) {
+	proxyURL1, _ := url.Parse("http://peer1.example.com:8080")
+	proxyURL2, _ := url.Parse("http://peer2.example.com:8080")
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    "http://peer1.example.com:8080",
+			ProxyURL: proxyURL1,
+			Models:   []string{"model-a", "model-b"},
+		},
+		"peer2": config.PeerConfig{
+			Proxy:    "http://peer2.example.com:8080",
+			ProxyURL: proxyURL2,
+			Models:   []string{"model-c", "model-d"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+	assert.Len(t, pm.proxyMap, 4)
+	assert.True(t, pm.HasPeerModel("model-a"))
+	assert.True(t, pm.HasPeerModel("model-b"))
+	assert.True(t, pm.HasPeerModel("model-c"))
+	assert.True(t, pm.HasPeerModel("model-d"))
+}
+
+func TestNewPeerProxy_DuplicateModelWarning(t *testing.T) {
+	// When the same model is in multiple peers, only the first (lexicographically by peer ID)
+	// should be mapped, and a warning should be logged
+	proxyURL1, _ := url.Parse("http://peer1.example.com:8080")
+	proxyURL2, _ := url.Parse("http://peer2.example.com:8080")
+	peers := config.PeerDictionaryConfig{
+		"alpha-peer": config.PeerConfig{
+			Proxy:    "http://peer1.example.com:8080",
+			ProxyURL: proxyURL1,
+			Models:   []string{"duplicate-model"},
+		},
+		"beta-peer": config.PeerConfig{
+			Proxy:    "http://peer2.example.com:8080",
+			ProxyURL: proxyURL2,
+			Models:   []string{"duplicate-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+	// Should only have one entry for the duplicate model
+	assert.Len(t, pm.proxyMap, 1)
+	assert.True(t, pm.HasPeerModel("duplicate-model"))
+}
+
+func TestHasPeerModel(t *testing.T) {
+	proxyURL, _ := url.Parse("http://peer1.example.com:8080")
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    "http://peer1.example.com:8080",
+			ProxyURL: proxyURL,
+			Models:   []string{"existing-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	assert.True(t, pm.HasPeerModel("existing-model"))
+	assert.False(t, pm.HasPeerModel("non-existing-model"))
+}
+
+func TestProxyRequest_ModelNotFound(t *testing.T) {
+	peers := config.PeerDictionaryConfig{}
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("non-existing-model", w, req)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "no peer proxy found for model non-existing-model")
+}
+
+func TestProxyRequest_Success(t *testing.T) {
+	// Create a test server to act as the peer
+	testServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusOK)
+		w.Write([]byte("response from peer"))
+	}))
+	defer testServer.Close()
+
+	proxyURL, _ := url.Parse(testServer.URL)
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    testServer.URL,
+			ProxyURL: proxyURL,
+			Models:   []string{"test-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("test-model", w, req)
+	assert.NoError(t, err)
+	assert.Equal(t, http.StatusOK, w.Code)
+	assert.Equal(t, "response from peer", w.Body.String())
+}
+
+func TestProxyRequest_ApiKeyInjection(t *testing.T) {
+	// Create a test server that checks for the Authorization header
+	var receivedAuthHeader string
+	testServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		receivedAuthHeader = r.Header.Get("Authorization")
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer testServer.Close()
+
+	proxyURL, _ := url.Parse(testServer.URL)
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    testServer.URL,
+			ProxyURL: proxyURL,
+			ApiKey:   "secret-api-key",
+			Models:   []string{"test-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("test-model", w, req)
+	assert.NoError(t, err)
+	assert.Equal(t, "Bearer secret-api-key", receivedAuthHeader)
+}
+
+func TestProxyRequest_NoApiKey(t *testing.T) {
+	// Create a test server that checks for the Authorization header
+	var receivedAuthHeader string
+	testServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		receivedAuthHeader = r.Header.Get("Authorization")
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer testServer.Close()
+
+	proxyURL, _ := url.Parse(testServer.URL)
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    testServer.URL,
+			ProxyURL: proxyURL,
+			ApiKey:   "", // No API key
+			Models:   []string{"test-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("test-model", w, req)
+	assert.NoError(t, err)
+	assert.Empty(t, receivedAuthHeader)
+}
+
+func TestProxyRequest_HostHeaderSet(t *testing.T) {
+	// Create a test server that checks the Host header
+	var receivedHost string
+	testServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		receivedHost = r.Host
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer testServer.Close()
+
+	proxyURL, _ := url.Parse(testServer.URL)
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    testServer.URL,
+			ProxyURL: proxyURL,
+			Models:   []string{"test-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("test-model", w, req)
+	assert.NoError(t, err)
+	// The Host header should be set to the target URL's host
+	assert.True(t, strings.HasPrefix(receivedHost, "127.0.0.1:"))
+}
+
+func TestProxyRequest_SSEHeaderModification(t *testing.T) {
+	// Create a test server that returns SSE content type
+	testServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/event-stream")
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer testServer.Close()
+
+	proxyURL, _ := url.Parse(testServer.URL)
+	peers := config.PeerDictionaryConfig{
+		"peer1": config.PeerConfig{
+			Proxy:    testServer.URL,
+			ProxyURL: proxyURL,
+			Models:   []string{"test-model"},
+		},
+	}
+
+	pm, err := NewPeerProxy(peers, testLogger)
+	require.NoError(t, err)
+
+	req := httptest.NewRequest("POST", "/v1/chat/completions", nil)
+	w := httptest.NewRecorder()
+
+	err = pm.ProxyRequest("test-model", w, req)
+	assert.NoError(t, err)
+	// The X-Accel-Buffering header should be set to "no" for SSE
+	assert.Equal(t, "no", w.Header().Get("X-Accel-Buffering"))
+}

proxy/process_test.go

@@ -395,6 +395,10 @@ func TestProcess_StopImmediately(t *testing.T) {
 // Test that SIGKILL is sent when gracefulStopTimeout is reached and properly terminates
 // the upstream command
 func TestProcess_ForceStopWithKill(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping slow test")
+	}
+
 	if runtime.GOOS == "windows" {
 		t.Skip("skipping SIGTERM test on Windows ")
 	}

proxy/processgroup_test.go

@@ -49,6 +49,10 @@ func TestProcessGroup_HasMember(t *testing.T) {
 // TestProcessGroup_ProxyRequestSwapIsTrueParallel tests that when swap is true
 // and multiple requests are made in parallel, only one process is running at a time.
 func TestProcessGroup_ProxyRequestSwapIsTrueParallel(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping slow test")
+	}
+
 	var processGroupTestConfig = config.AddDefaultGroupToConfig(config.Config{
 		HealthCheckTimeout: 15,
 		Models: map[string]config.ModelConfig{

proxy/proxymanager.go

@@ -3,6 +3,7 @@ package proxy
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"fmt"
 	"io"
 	"mime/multipart"
@@ -50,19 +51,42 @@ type ProxyManager struct {
 	buildDate string
 	commit    string
 	version   string
+
+	// peer proxy see: #296, #433
+	peerProxy *PeerProxy
 }
 
-func New(config config.Config) *ProxyManager {
+func New(proxyConfig config.Config) *ProxyManager {
 	// set up loggers
-	stdoutLogger := NewLogMonitorWriter(os.Stdout)
-	upstreamLogger := NewLogMonitorWriter(stdoutLogger)
-	proxyLogger := NewLogMonitorWriter(stdoutLogger)
 
-	if config.LogRequests {
+	var muxLogger, upstreamLogger, proxyLogger *LogMonitor
+	switch proxyConfig.LogToStdout {
+	case config.LogToStdoutNone:
+		muxLogger = NewLogMonitorWriter(io.Discard)
+		upstreamLogger = NewLogMonitorWriter(io.Discard)
+		proxyLogger = NewLogMonitorWriter(io.Discard)
+	case config.LogToStdoutBoth:
+		muxLogger = NewLogMonitorWriter(os.Stdout)
+		upstreamLogger = NewLogMonitorWriter(muxLogger)
+		proxyLogger = NewLogMonitorWriter(muxLogger)
+	case config.LogToStdoutUpstream:
+		muxLogger = NewLogMonitorWriter(os.Stdout)
+		upstreamLogger = NewLogMonitorWriter(muxLogger)
+		proxyLogger = NewLogMonitorWriter(io.Discard)
+	default:
+		// same as config.LogToStdoutProxy
+		// helpful because some old tests create a config.Config directly and it
+		// may not have LogToStdout set explicitly
+		muxLogger = NewLogMonitorWriter(os.Stdout)
+		upstreamLogger = NewLogMonitorWriter(io.Discard)
+		proxyLogger = NewLogMonitorWriter(muxLogger)
+	}
+
+	if proxyConfig.LogRequests {
 		proxyLogger.Warn("LogRequests configuration is deprecated. Use logLevel instead.")
 	}
 
-	switch strings.ToLower(strings.TrimSpace(config.LogLevel)) {
+	switch strings.ToLower(strings.TrimSpace(proxyConfig.LogLevel)) {
 	case "debug":
 		proxyLogger.SetLogLevel(LevelDebug)
 		upstreamLogger.SetLogLevel(LevelDebug)
@@ -99,7 +123,7 @@ func New(config config.Config) *ProxyManager {
 		"stampnano":   time.StampNano,
 	}
 
-	if timeFormat, ok := timeFormats[strings.ToLower(strings.TrimSpace(config.LogTimeFormat))]; ok {
+	if timeFormat, ok := timeFormats[strings.ToLower(strings.TrimSpace(proxyConfig.LogTimeFormat))]; ok {
 		proxyLogger.SetLogTimeFormat(timeFormat)
 		upstreamLogger.SetLogTimeFormat(timeFormat)
 	}
@@ -107,18 +131,24 @@ func New(config config.Config) *ProxyManager {
 	shutdownCtx, shutdownCancel := context.WithCancel(context.Background())
 
 	var maxMetrics int
-	if config.MetricsMaxInMemory <= 0 {
+	if proxyConfig.MetricsMaxInMemory <= 0 {
 		maxMetrics = 1000 // Default fallback
 	} else {
-		maxMetrics = config.MetricsMaxInMemory
+		maxMetrics = proxyConfig.MetricsMaxInMemory
+	}
+
+	peerProxy, err := NewPeerProxy(proxyConfig.Peers, proxyLogger)
+	if err != nil {
+		proxyLogger.Errorf("Disabling Peering. Failed to create proxy peers: %v", err)
+		peerProxy = nil
 	}
 
 	pm := &ProxyManager{
-		config:    config,
+		config:    proxyConfig,
 		ginEngine: gin.New(),
 
 		proxyLogger:    proxyLogger,
-		muxLogger:      stdoutLogger,
+		muxLogger:      muxLogger,
 		upstreamLogger: upstreamLogger,
 
 		metricsMonitor: newMetricsMonitor(proxyLogger, maxMetrics),
@@ -131,37 +161,46 @@ func New(config config.Config) *ProxyManager {
 		buildDate: "unknown",
 		commit:    "abcd1234",
 		version:   "0",
+
+		peerProxy: peerProxy,
 	}
 
 	// create the process groups
-	for groupID := range config.Groups {
-		processGroup := NewProcessGroup(groupID, config, proxyLogger, upstreamLogger)
+	for groupID := range proxyConfig.Groups {
+		processGroup := NewProcessGroup(groupID, proxyConfig, proxyLogger, upstreamLogger)
 		pm.processGroups[groupID] = processGroup
 	}
 
 	pm.setupGinEngine()
 
 	// run any startup hooks
-	if len(config.Hooks.OnStartup.Preload) > 0 {
+	if len(proxyConfig.Hooks.OnStartup.Preload) > 0 {
 		// do it in the background, don't block startup -- not sure if good idea yet
 		go func() {
 			discardWriter := &DiscardWriter{}
-			for _, realModelName := range config.Hooks.OnStartup.Preload {
-				proxyLogger.Infof("Preloading model: %s", realModelName)
-				processGroup, _, err := pm.swapProcessGroup(realModelName)
+			for _, preloadModelName := range proxyConfig.Hooks.OnStartup.Preload {
+				modelID, ok := proxyConfig.RealModelName(preloadModelName)
+
+				if !ok {
+					proxyLogger.Warnf("Preload model %s not found in config", preloadModelName)
+					continue
+				}
+
+				proxyLogger.Infof("Preloading model: %s", modelID)
+				processGroup, err := pm.swapProcessGroup(modelID)
 
 				if err != nil {
 					event.Emit(ModelPreloadedEvent{
-						ModelName: realModelName,
+						ModelName: modelID,
 						Success:   false,
 					})
-					proxyLogger.Errorf("Failed to preload model %s: %v", realModelName, err)
+					proxyLogger.Errorf("Failed to preload model %s: %v", modelID, err)
 					continue
 				} else {
 					req, _ := http.NewRequest("GET", "/", nil)
-					processGroup.ProxyRequest(realModelName, discardWriter, req)
+					processGroup.ProxyRequest(modelID, discardWriter, req)
 					event.Emit(ModelPreloadedEvent{
-						ModelName: realModelName,
+						ModelName: modelID,
 						Success:   true,
 					})
 				}
@@ -236,37 +275,41 @@ func (pm *ProxyManager) setupGinEngine() {
 	})
 
 	// Set up routes using the Gin engine
-	pm.ginEngine.POST("/v1/chat/completions", pm.proxyInferenceHandler)
+	// Protected routes use pm.apiKeyAuth() middleware
+	pm.ginEngine.POST("/v1/chat/completions", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 	// Support legacy /v1/completions api, see issue #12
-	pm.ginEngine.POST("/v1/completions", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/completions", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 	// Support anthropic /v1/messages (added https://github.com/ggml-org/llama.cpp/pull/17570)
-	pm.ginEngine.POST("/v1/messages", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/messages", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 
 	// Support embeddings and reranking
-	pm.ginEngine.POST("/v1/embeddings", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/embeddings", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 
 	// llama-server's /reranking endpoint + aliases
-	pm.ginEngine.POST("/reranking", pm.proxyInferenceHandler)
-	pm.ginEngine.POST("/rerank", pm.proxyInferenceHandler)
-	pm.ginEngine.POST("/v1/rerank", pm.proxyInferenceHandler)
-	pm.ginEngine.POST("/v1/reranking", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/reranking", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/rerank", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/rerank", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/reranking", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 
 	// llama-server's /infill endpoint for code infilling
-	pm.ginEngine.POST("/infill", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/infill", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 
 	// llama-server's /completion endpoint
-	pm.ginEngine.POST("/completion", pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/completion", pm.apiKeyAuth(), pm.proxyInferenceHandler)
 
 	// Support audio/speech endpoint
-	pm.ginEngine.POST("/v1/audio/speech", pm.proxyInferenceHandler)
-	pm.ginEngine.POST("/v1/audio/transcriptions", pm.proxyOAIPostFormHandler)
+	pm.ginEngine.POST("/v1/audio/speech", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/audio/voices", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/audio/transcriptions", pm.apiKeyAuth(), pm.proxyOAIPostFormHandler)
+	pm.ginEngine.POST("/v1/images/generations", pm.apiKeyAuth(), pm.proxyInferenceHandler)
+	pm.ginEngine.POST("/v1/images/edits", pm.apiKeyAuth(), pm.proxyOAIPostFormHandler)
 
-	pm.ginEngine.GET("/v1/models", pm.listModelsHandler)
+	pm.ginEngine.GET("/v1/models", pm.apiKeyAuth(), pm.listModelsHandler)
 
 	// in proxymanager_loghandlers.go
-	pm.ginEngine.GET("/logs", pm.sendLogsHandlers)
-	pm.ginEngine.GET("/logs/stream", pm.streamLogsHandler)
-	pm.ginEngine.GET("/logs/stream/:logMonitorID", pm.streamLogsHandler)
+	pm.ginEngine.GET("/logs", pm.apiKeyAuth(), pm.sendLogsHandlers)
+	pm.ginEngine.GET("/logs/stream", pm.apiKeyAuth(), pm.streamLogsHandler)
+	pm.ginEngine.GET("/logs/stream/*logMonitorID", pm.apiKeyAuth(), pm.streamLogsHandler)
 
 	/**
 	 * User Interface Endpoints
@@ -278,9 +321,9 @@ func (pm *ProxyManager) setupGinEngine() {
 	pm.ginEngine.GET("/upstream", func(c *gin.Context) {
 		c.Redirect(http.StatusFound, "/ui/models")
 	})
-	pm.ginEngine.Any("/upstream/*upstreamPath", pm.proxyToUpstream)
-	pm.ginEngine.GET("/unload", pm.unloadAllModelsHandler)
-	pm.ginEngine.GET("/running", pm.listRunningProcessesHandler)
+	pm.ginEngine.Any("/upstream/*upstreamPath", pm.apiKeyAuth(), pm.proxyToUpstream)
+	pm.ginEngine.GET("/unload", pm.apiKeyAuth(), pm.unloadAllModelsHandler)
+	pm.ginEngine.GET("/running", pm.apiKeyAuth(), pm.listRunningProcessesHandler)
 	pm.ginEngine.GET("/health", func(c *gin.Context) {
 		c.String(http.StatusOK, "OK")
 	})
@@ -378,16 +421,10 @@ func (pm *ProxyManager) Shutdown() {
 	pm.shutdownCancel()
 }
 
-func (pm *ProxyManager) swapProcessGroup(requestedModel string) (*ProcessGroup, string, error) {
-	// de-alias the real model name and get a real one
-	realModelName, found := pm.config.RealModelName(requestedModel)
-	if !found {
-		return nil, realModelName, fmt.Errorf("could not find real modelID for %s", requestedModel)
-	}
-
+func (pm *ProxyManager) swapProcessGroup(realModelName string) (*ProcessGroup, error) {
 	processGroup := pm.findGroupByModelName(realModelName)
 	if processGroup == nil {
-		return nil, realModelName, fmt.Errorf("could not find process group for model %s", requestedModel)
+		return nil, fmt.Errorf("could not find process group for model %s", realModelName)
 	}
 
 	if processGroup.exclusive {
@@ -399,54 +436,71 @@ func (pm *ProxyManager) swapProcessGroup(requestedModel string) (*ProcessGroup,
 		}
 	}
 
-	return processGroup, realModelName, nil
+	return processGroup, nil
 }
 
 func (pm *ProxyManager) listModelsHandler(c *gin.Context) {
 	data := make([]gin.H, 0, len(pm.config.Models))
 	createdTime := time.Now().Unix()
 
+	newRecord := func(modelId string, modelConfig config.ModelConfig) gin.H {
+		record := gin.H{
+			"id":       modelId,
+			"object":   "model",
+			"created":  createdTime,
+			"owned_by": "llama-swap",
+		}
+
+		if name := strings.TrimSpace(modelConfig.Name); name != "" {
+			record["name"] = name
+		}
+		if desc := strings.TrimSpace(modelConfig.Description); desc != "" {
+			record["description"] = desc
+		}
+
+		// Add metadata if present
+		if len(modelConfig.Metadata) > 0 {
+			record["meta"] = gin.H{
+				"llamaswap": modelConfig.Metadata,
+			}
+		}
+		return record
+	}
+
 	for id, modelConfig := range pm.config.Models {
 		if modelConfig.Unlisted {
 			continue
 		}
 
-		newRecord := func(modelId string) gin.H {
-			record := gin.H{
-				"id":       modelId,
-				"object":   "model",
-				"created":  createdTime,
-				"owned_by": "llama-swap",
-			}
-
-			if name := strings.TrimSpace(modelConfig.Name); name != "" {
-				record["name"] = name
-			}
-			if desc := strings.TrimSpace(modelConfig.Description); desc != "" {
-				record["description"] = desc
-			}
-
-			// Add metadata if present
-			if len(modelConfig.Metadata) > 0 {
-				record["meta"] = gin.H{
-					"llamaswap": modelConfig.Metadata,
-				}
-			}
-			return record
-		}
-
-		data = append(data, newRecord(id))
+		data = append(data, newRecord(id, modelConfig))
 
 		// Include aliases
 		if pm.config.IncludeAliasesInList {
 			for _, alias := range modelConfig.Aliases {
 				if alias := strings.TrimSpace(alias); alias != "" {
-					data = append(data, newRecord(alias))
+					data = append(data, newRecord(alias, modelConfig))
 				}
 			}
 		}
 	}
 
+	if pm.peerProxy != nil {
+		for peerID, peer := range pm.peerProxy.ListPeers() {
+			// add peer models
+			for _, modelID := range peer.Models {
+				// Skip unlisted models if not showing them
+				record := newRecord(modelID, config.ModelConfig{
+					Name: fmt.Sprintf("%s: %s", peerID, modelID),
+					Metadata: map[string]any{
+						"peerID": peerID,
+					},
+				})
+
+				data = append(data, record)
+			}
+		}
+	}
+
 	// Sort by the "id" key
 	sort.Slice(data, func(i, j int) bool {
 		si, _ := data[i]["id"].(string)
@@ -466,62 +520,61 @@ func (pm *ProxyManager) listModelsHandler(c *gin.Context) {
 	})
 }
 
-func (pm *ProxyManager) proxyToUpstream(c *gin.Context) {
-	upstreamPath := c.Param("upstreamPath")
-
-	// split the upstream path by / and search for the model name
-	parts := strings.Split(strings.TrimSpace(upstreamPath), "/")
-	if len(parts) == 0 {
-		pm.sendErrorResponse(c, http.StatusBadRequest, "model id required in path")
-		return
-	}
-
-	modelFound := false
+// findModelInPath searches for a valid model name in a path with slashes.
+// It iteratively builds up path segments until it finds a matching model.
+// Returns: (searchModelName, realModelName, remainingPath, found)
+// Example: "/author/model/endpoint" with model "author/model" -> ("author/model", "author/model", "/endpoint", true)
+func (pm *ProxyManager) findModelInPath(path string) (searchName string, realName string, remainingPath string, found bool) {
+	parts := strings.Split(strings.TrimSpace(path), "/")
 	searchModelName := ""
-	var modelName, remainingPath string
+
 	for i, part := range parts {
-		if parts[i] == "" {
+		if part == "" {
 			continue
 		}
 
 		if searchModelName == "" {
 			searchModelName = part
 		} else {
-			searchModelName = searchModelName + "/" + parts[i]
+			searchModelName = searchModelName + "/" + part
 		}
 
-		if real, ok := pm.config.RealModelName(searchModelName); ok {
-			modelName = real
-			remainingPath = "/" + strings.Join(parts[i+1:], "/")
-			modelFound = true
-
-			// Check if this is exactly a model name with no additional path
-			// and doesn't end with a trailing slash
-			if remainingPath == "/" && !strings.HasSuffix(upstreamPath, "/") {
-				// Build new URL with query parameters preserved
-				newPath := "/upstream/" + searchModelName + "/"
-				if c.Request.URL.RawQuery != "" {
-					newPath += "?" + c.Request.URL.RawQuery
-				}
-
-				// Use 308 for non-GET/HEAD requests to preserve method
-				if c.Request.Method == http.MethodGet || c.Request.Method == http.MethodHead {
-					c.Redirect(http.StatusMovedPermanently, newPath)
-				} else {
-					c.Redirect(http.StatusPermanentRedirect, newPath)
-				}
-				return
-			}
-			break
+		if modelID, ok := pm.config.RealModelName(searchModelName); ok {
+			return searchModelName, modelID, "/" + strings.Join(parts[i+1:], "/"), true
 		}
 	}
 
+	return "", "", "", false
+}
+
+func (pm *ProxyManager) proxyToUpstream(c *gin.Context) {
+	upstreamPath := c.Param("upstreamPath")
+
+	searchModelName, modelID, remainingPath, modelFound := pm.findModelInPath(upstreamPath)
+
 	if !modelFound {
 		pm.sendErrorResponse(c, http.StatusBadRequest, "model id required in path")
 		return
 	}
 
-	processGroup, realModelName, err := pm.swapProcessGroup(modelName)
+	// Redirect /upstream/modelname to /upstream/modelname/ for URL consistency.
+	// This ensures relative URLs in upstream responses resolve correctly and
+	// provides canonical URL form. Uses 308 for POST/PUT/etc to preserve the
+	// HTTP method (301 would downgrade to GET).
+	if remainingPath == "/" && !strings.HasSuffix(upstreamPath, "/") {
+		newPath := "/upstream/" + searchModelName + "/"
+		if c.Request.URL.RawQuery != "" {
+			newPath += "?" + c.Request.URL.RawQuery
+		}
+		if c.Request.Method == http.MethodGet || c.Request.Method == http.MethodHead {
+			c.Redirect(http.StatusMovedPermanently, newPath)
+		} else {
+			c.Redirect(http.StatusPermanentRedirect, newPath)
+		}
+		return
+	}
+
+	processGroup, err := pm.swapProcessGroup(modelID)
 	if err != nil {
 		pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error swapping process group: %s", err.Error()))
 		return
@@ -533,15 +586,15 @@ func (pm *ProxyManager) proxyToUpstream(c *gin.Context) {
 
 	// attempt to record metrics if it is a POST request
 	if pm.metricsMonitor != nil && c.Request.Method == "POST" {
-		if err := pm.metricsMonitor.wrapHandler(realModelName, c.Writer, c.Request, processGroup.ProxyRequest); err != nil {
+		if err := pm.metricsMonitor.wrapHandler(modelID, c.Writer, c.Request, processGroup.ProxyRequest); err != nil {
 			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error proxying metrics wrapped request: %s", err.Error()))
-			pm.proxyLogger.Errorf("Error proxying wrapped upstream request for model %s, path=%s", realModelName, originalPath)
+			pm.proxyLogger.Errorf("Error proxying wrapped upstream request for model %s, path=%s", modelID, originalPath)
 			return
 		}
 	} else {
-		if err := processGroup.ProxyRequest(realModelName, c.Writer, c.Request); err != nil {
+		if err := processGroup.ProxyRequest(modelID, c.Writer, c.Request); err != nil {
 			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error proxying request: %s", err.Error()))
-			pm.proxyLogger.Errorf("Error proxying upstream request for model %s, path=%s", realModelName, originalPath)
+			pm.proxyLogger.Errorf("Error proxying upstream request for model %s, path=%s", modelID, originalPath)
 			return
 		}
 	}
@@ -560,41 +613,54 @@ func (pm *ProxyManager) proxyInferenceHandler(c *gin.Context) {
 		return
 	}
 
-	realModelName, found := pm.config.RealModelName(requestedModel)
-	if !found {
-		pm.sendErrorResponse(c, http.StatusBadRequest, fmt.Sprintf("could not find real modelID for %s", requestedModel))
-		return
-	}
+	// Look for a matching local model first
+	var nextHandler func(modelID string, w http.ResponseWriter, r *http.Request) error
 
-	processGroup, _, err := pm.swapProcessGroup(realModelName)
-	if err != nil {
-		pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error swapping process group: %s", err.Error()))
-		return
-	}
-
-	// issue #69 allow custom model names to be sent to upstream
-	useModelName := pm.config.Models[realModelName].UseModelName
-	if useModelName != "" {
-		bodyBytes, err = sjson.SetBytes(bodyBytes, "model", useModelName)
+	modelID, found := pm.config.RealModelName(requestedModel)
+	if found {
+		processGroup, err := pm.swapProcessGroup(modelID)
 		if err != nil {
-			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error rewriting model name in JSON: %s", err.Error()))
+			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error swapping process group: %s", err.Error()))
 			return
 		}
-	}
 
-	// issue #174 strip parameters from the JSON body
-	stripParams, err := pm.config.Models[realModelName].Filters.SanitizedStripParams()
-	if err != nil { // just log it and continue
-		pm.proxyLogger.Errorf("Error sanitizing strip params string: %s, %s", pm.config.Models[realModelName].Filters.StripParams, err.Error())
-	} else {
-		for _, param := range stripParams {
-			pm.proxyLogger.Debugf("<%s> stripping param: %s", realModelName, param)
-			bodyBytes, err = sjson.DeleteBytes(bodyBytes, param)
+		// issue #69 allow custom model names to be sent to upstream
+		useModelName := pm.config.Models[modelID].UseModelName
+		if useModelName != "" {
+			bodyBytes, err = sjson.SetBytes(bodyBytes, "model", useModelName)
 			if err != nil {
-				pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error deleting parameter %s from request", param))
+				pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error rewriting model name in JSON: %s", err.Error()))
 				return
 			}
 		}
+
+		// issue #174 strip parameters from the JSON body
+		stripParams, err := pm.config.Models[modelID].Filters.SanitizedStripParams()
+		if err != nil { // just log it and continue
+			pm.proxyLogger.Errorf("Error sanitizing strip params string: %s, %s", pm.config.Models[modelID].Filters.StripParams, err.Error())
+		} else {
+			for _, param := range stripParams {
+				pm.proxyLogger.Debugf("<%s> stripping param: %s", modelID, param)
+				bodyBytes, err = sjson.DeleteBytes(bodyBytes, param)
+				if err != nil {
+					pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error deleting parameter %s from request", param))
+					return
+				}
+			}
+		}
+
+		pm.proxyLogger.Debugf("ProxyManager using local Process for model: %s", requestedModel)
+		nextHandler = processGroup.ProxyRequest
+	} else if pm.peerProxy != nil && pm.peerProxy.HasPeerModel(requestedModel) {
+		pm.proxyLogger.Debugf("ProxyManager using ProxyPeer for model: %s", requestedModel)
+		modelID = requestedModel
+		nextHandler = pm.peerProxy.ProxyRequest
+
+	}
+
+	if nextHandler == nil {
+		pm.sendErrorResponse(c, http.StatusBadRequest, fmt.Sprintf("could not find suitable inference handler for %s", requestedModel))
+		return
 	}
 
 	c.Request.Body = io.NopCloser(bytes.NewBuffer(bodyBytes))
@@ -607,19 +673,19 @@ func (pm *ProxyManager) proxyInferenceHandler(c *gin.Context) {
 	// issue #366 extract values that downstream handlers may need
 	isStreaming := gjson.GetBytes(bodyBytes, "stream").Bool()
 	ctx := context.WithValue(c.Request.Context(), proxyCtxKey("streaming"), isStreaming)
-	ctx = context.WithValue(ctx, proxyCtxKey("model"), realModelName)
+	ctx = context.WithValue(ctx, proxyCtxKey("model"), modelID)
 	c.Request = c.Request.WithContext(ctx)
 
 	if pm.metricsMonitor != nil && c.Request.Method == "POST" {
-		if err := pm.metricsMonitor.wrapHandler(realModelName, c.Writer, c.Request, processGroup.ProxyRequest); err != nil {
+		if err := pm.metricsMonitor.wrapHandler(modelID, c.Writer, c.Request, nextHandler); err != nil {
 			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error proxying metrics wrapped request: %s", err.Error()))
-			pm.proxyLogger.Errorf("Error Proxying Metrics Wrapped Request for processGroup %s and model %s", processGroup.id, realModelName)
+			pm.proxyLogger.Errorf("Error Proxying Metrics Wrapped Request model %s", modelID)
 			return
 		}
 	} else {
-		if err := processGroup.ProxyRequest(realModelName, c.Writer, c.Request); err != nil {
+		if err := nextHandler(modelID, c.Writer, c.Request); err != nil {
 			pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error proxying request: %s", err.Error()))
-			pm.proxyLogger.Errorf("Error Proxying Request for processGroup %s and model %s", processGroup.id, realModelName)
+			pm.proxyLogger.Errorf("Error Proxying Request for model %s", modelID)
 			return
 		}
 	}
@@ -639,7 +705,13 @@ func (pm *ProxyManager) proxyOAIPostFormHandler(c *gin.Context) {
 		return
 	}
 
-	processGroup, realModelName, err := pm.swapProcessGroup(requestedModel)
+	modelID, found := pm.config.RealModelName(requestedModel)
+	if !found {
+		pm.sendErrorResponse(c, http.StatusBadRequest, fmt.Sprintf("could not find real modelID for %s", requestedModel))
+		return
+	}
+
+	processGroup, err := pm.swapProcessGroup(modelID)
 	if err != nil {
 		pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error swapping process group: %s", err.Error()))
 		return
@@ -657,7 +729,7 @@ func (pm *ProxyManager) proxyOAIPostFormHandler(c *gin.Context) {
 			// If this is the model field and we have a profile, use just the model name
 			if key == "model" {
 				// # issue #69 allow custom model names to be sent to upstream
-				useModelName := pm.config.Models[realModelName].UseModelName
+				useModelName := pm.config.Models[modelID].UseModelName
 
 				if useModelName != "" {
 					fieldValue = useModelName
@@ -728,9 +800,9 @@ func (pm *ProxyManager) proxyOAIPostFormHandler(c *gin.Context) {
 	modifiedReq.ContentLength = int64(requestBuffer.Len())
 
 	// Use the modified request for proxying
-	if err := processGroup.ProxyRequest(realModelName, c.Writer, modifiedReq); err != nil {
+	if err := processGroup.ProxyRequest(modelID, c.Writer, modifiedReq); err != nil {
 		pm.sendErrorResponse(c, http.StatusInternalServerError, fmt.Sprintf("error proxying request: %s", err.Error()))
-		pm.proxyLogger.Errorf("Error Proxying Request for processGroup %s and model %s", processGroup.id, realModelName)
+		pm.proxyLogger.Errorf("Error Proxying Request for processGroup %s and model %s", processGroup.id, modelID)
 		return
 	}
 }
@@ -745,6 +817,67 @@ func (pm *ProxyManager) sendErrorResponse(c *gin.Context, statusCode int, messag
 	}
 }
 
+// apiKeyAuth returns a middleware that validates API keys if configured.
+// Returns a pass-through handler if no API keys are configured.
+func (pm *ProxyManager) apiKeyAuth() gin.HandlerFunc {
+	if len(pm.config.RequiredAPIKeys) == 0 {
+		return func(c *gin.Context) { c.Next() }
+	}
+
+	return func(c *gin.Context) {
+		xApiKey := c.GetHeader("x-api-key")
+
+		var bearerKey string
+		var basicKey string
+		if auth := c.GetHeader("Authorization"); auth != "" {
+			if strings.HasPrefix(auth, "Bearer ") {
+				bearerKey = strings.TrimPrefix(auth, "Bearer ")
+			} else if strings.HasPrefix(auth, "Basic ") {
+				// Basic Auth: base64(username:password), password is the API key
+				encoded := strings.TrimPrefix(auth, "Basic ")
+				if decoded, err := base64.StdEncoding.DecodeString(encoded); err == nil {
+					parts := strings.SplitN(string(decoded), ":", 2)
+					if len(parts) == 2 {
+						basicKey = parts[1] // password is the API key
+					}
+				}
+			}
+		}
+
+		// Use first key found: Basic, then Bearer, then x-api-key
+		var providedKey string
+		if basicKey != "" {
+			providedKey = basicKey
+		} else if bearerKey != "" {
+			providedKey = bearerKey
+		} else {
+			providedKey = xApiKey
+		}
+
+		// Validate key
+		valid := false
+		for _, key := range pm.config.RequiredAPIKeys {
+			if providedKey == key {
+				valid = true
+				break
+			}
+		}
+
+		if !valid {
+			c.Header("WWW-Authenticate", `Basic realm="llama-swap"`)
+			pm.sendErrorResponse(c, http.StatusUnauthorized, "unauthorized: invalid or missing API key")
+			c.Abort()
+			return
+		}
+
+		// Strip auth headers to prevent leakage to upstream
+		c.Request.Header.Del("Authorization")
+		c.Request.Header.Del("x-api-key")
+
+		c.Next()
+	}
+}
+
 func (pm *ProxyManager) unloadAllModelsHandler(c *gin.Context) {
 	pm.StopProcesses(StopImmediately)
 	c.String(http.StatusOK, "OK")

proxy/proxymanager_api.go

@@ -18,11 +18,13 @@ type Model struct {
 	Description string `json:"description"`
 	State       string `json:"state"`
 	Unlisted    bool   `json:"unlisted"`
+	PeerID      string `json:"peerID"`
 }
 
 func addApiHandlers(pm *ProxyManager) {
 	// Add API endpoints for React to consume
-	apiGroup := pm.ginEngine.Group("/api")
+	// Protected with API key authentication
+	apiGroup := pm.ginEngine.Group("/api", pm.apiKeyAuth())
 	{
 		apiGroup.POST("/models/unload", pm.apiUnloadAllModels)
 		apiGroup.POST("/models/unload/*model", pm.apiUnloadSingleModelHandler)
@@ -82,6 +84,18 @@ func (pm *ProxyManager) getModelStatus() []Model {
 		})
 	}
 
+	// Iterate over the peer models
+	if pm.peerProxy != nil {
+		for peerID, peer := range pm.peerProxy.ListPeers() {
+			for _, modelID := range peer.Models {
+				models = append(models, Model{
+					Id:     modelID,
+					PeerID: peerID,
+				})
+			}
+		}
+	}
+
 	return models
 }
 

proxy/proxymanager_loghandlers.go

@@ -31,7 +31,7 @@ func (pm *ProxyManager) streamLogsHandler(c *gin.Context) {
 	// prevent nginx from buffering streamed logs
 	c.Header("X-Accel-Buffering", "no")
 
-	logMonitorId := c.Param("logMonitorID")
+	logMonitorId := strings.TrimPrefix(c.Param("logMonitorID"), "/")
 	logger, err := pm.getLogger(logMonitorId)
 	if err != nil {
 		c.String(http.StatusBadRequest, err.Error())
@@ -92,8 +92,9 @@ func (pm *ProxyManager) getLogger(logMonitorId string) (*LogMonitor, error) {
 	case "upstream":
 		return pm.upstreamLogger, nil
 	default:
-		// search for a models specific logger
-		if name, found := pm.config.RealModelName(logMonitorId); found {
+		// search for a models specific logger using findModelInPath
+		// to handle model names with slashes (e.g., "author/model")
+		if _, name, _, found := pm.findModelInPath("/" + logMonitorId); found {
 			for _, group := range pm.processGroups {
 				if process, found := group.GetMember(name); found {
 					return process.Logger(), nil
@@ -101,6 +102,6 @@ func (pm *ProxyManager) getLogger(logMonitorId string) (*LogMonitor, error) {
 			}
 		}
 
-		return nil, fmt.Errorf("invalid logger. Use 'proxy' or 'upstream'")
+		return nil, fmt.Errorf("invalid logger. Use 'proxy', 'upstream' or a model's ID")
 	}
 }

proxy/proxymanager_test.go

@@ -3,6 +3,7 @@ package proxy
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"encoding/json"
 	"fmt"
 	"math/rand"
@@ -36,10 +37,6 @@ func (r *TestResponseRecorder) CloseNotify() <-chan bool {
 	return r.closeChannel
 }
 
-func (r *TestResponseRecorder) closeClient() {
-	r.closeChannel <- true
-}
-
 func CreateTestResponseRecorder() *TestResponseRecorder {
 	return &TestResponseRecorder{
 		httptest.NewRecorder(),
@@ -223,17 +220,23 @@ func TestProxyManager_ListModelsHandler(t *testing.T) {
 	model2Config.Name = "     " // empty whitespace only strings will get ignored
 	model2Config.Description = "  "
 
-	config := config.Config{
+	cfg := config.Config{
 		HealthCheckTimeout: 15,
 		Models: map[string]config.ModelConfig{
 			"model1": model1Config,
 			"model2": model2Config,
 			"model3": getTestSimpleResponderConfig("model3"),
 		},
+		Peers: map[string]config.PeerConfig{
+			"peer1": {
+				Proxy:  "http://peer1:8080",
+				Models: []string{"peer-model-a", "peer-model-b"},
+			},
+		},
 		LogLevel: "error",
 	}
 
-	proxy := New(config)
+	proxy := New(cfg)
 
 	// Create a test request
 	req := httptest.NewRequest("GET", "/v1/models", nil)
@@ -258,14 +261,16 @@ func TestProxyManager_ListModelsHandler(t *testing.T) {
 		t.Fatalf("Failed to parse JSON response: %v", err)
 	}
 
-	// Check the number of models returned
-	assert.Len(t, response.Data, 3)
+	// Check the number of models returned (3 local + 2 peer models)
+	assert.Len(t, response.Data, 5)
 
 	// Check the details of each model
 	expectedModels := map[string]struct{}{
-		"model1": {},
-		"model2": {},
-		"model3": {},
+		"model1":       {},
+		"model2":       {},
+		"model3":       {},
+		"peer-model-a": {},
+		"peer-model-b": {},
 	}
 
 	// make all models
@@ -296,6 +301,19 @@ func TestProxyManager_ListModelsHandler(t *testing.T) {
 			description, ok := model["description"].(string)
 			assert.True(t, ok, "description should be a string")
 			assert.Equal(t, "Model 1 description is used for testing", description)
+		} else if modelID == "peer-model-a" || modelID == "peer-model-b" {
+			// Peer models should have meta.llamaswap.peerID
+			meta, exists := model["meta"]
+			assert.True(t, exists, "peer model should have meta field")
+			metaMap, ok := meta.(map[string]interface{})
+			assert.True(t, ok, "meta should be a map")
+			llamaswap, exists := metaMap["llamaswap"]
+			assert.True(t, exists, "meta should have llamaswap field")
+			llamaswapMap, ok := llamaswap.(map[string]interface{})
+			assert.True(t, ok, "llamaswap should be a map")
+			peerID, exists := llamaswapMap["peerID"]
+			assert.True(t, exists, "llamaswap should have peerID field")
+			assert.Equal(t, "peer1", peerID)
 		} else {
 			_, exists := model["name"]
 			assert.False(t, exists, "unexpected name field for model: %s", modelID)
@@ -502,6 +520,10 @@ func TestProxyManager_ListModelsHandler_IncludeAliasesInList(t *testing.T) {
 }
 
 func TestProxyManager_Shutdown(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping slow test")
+	}
+
 	// make broken model configurations
 	model1Config := getTestSimpleResponderConfigPort("model1", 9991)
 	model1Config.Proxy = "http://localhost:10001/"
@@ -1078,7 +1100,8 @@ func TestProxyManager_StreamingEndpointsReturnNoBufferingHeader(t *testing.T) {
 	config := config.AddDefaultGroupToConfig(config.Config{
 		HealthCheckTimeout: 15,
 		Models: map[string]config.ModelConfig{
-			"model1": getTestSimpleResponderConfig("model1"),
+			"model1":       getTestSimpleResponderConfig("model1"),
+			"author/model": getTestSimpleResponderConfig("author/model"),
 		},
 		LogLevel: "error",
 	})
@@ -1091,6 +1114,7 @@ func TestProxyManager_StreamingEndpointsReturnNoBufferingHeader(t *testing.T) {
 		"/logs/stream",
 		"/logs/stream/proxy",
 		"/logs/stream/upstream",
+		"/logs/stream/author/model",
 	}
 
 	for _, endpoint := range endpoints {
@@ -1185,3 +1209,349 @@ func TestProxyManager_ApiGetVersion(t *testing.T) {
 		assert.Equal(t, value, response[key], "%s value %s should match response %s", key, value, response[key])
 	}
 }
+
+func TestProxyManager_APIKeyAuth(t *testing.T) {
+	testConfig := config.AddDefaultGroupToConfig(config.Config{
+		HealthCheckTimeout: 15,
+		Models: map[string]config.ModelConfig{
+			"model1": getTestSimpleResponderConfig("model1"),
+		},
+		RequiredAPIKeys: []string{"valid-key-1", "valid-key-2"},
+		LogLevel:        "error",
+	})
+
+	proxy := New(testConfig)
+	defer proxy.StopProcesses(StopImmediately)
+
+	t.Run("valid key in x-api-key header", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		req.Header.Set("x-api-key", "valid-key-1")
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+
+	t.Run("valid key in Authorization Bearer header", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		req.Header.Set("Authorization", "Bearer valid-key-2")
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+
+	t.Run("both headers with matching keys", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		req.Header.Set("x-api-key", "valid-key-1")
+		req.Header.Set("Authorization", "Bearer valid-key-1")
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+
+	t.Run("invalid key returns 401", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		req.Header.Set("x-api-key", "invalid-key")
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusUnauthorized, w.Code)
+		assert.Contains(t, w.Body.String(), "unauthorized")
+	})
+
+	t.Run("missing key returns 401", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusUnauthorized, w.Code)
+	})
+
+	t.Run("valid key in Basic Auth header", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		// Basic Auth: base64("anyuser:valid-key-1")
+		credentials := base64.StdEncoding.EncodeToString([]byte("anyuser:valid-key-1"))
+		req.Header.Set("Authorization", "Basic "+credentials)
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+
+	t.Run("invalid key in Basic Auth header returns 401", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		credentials := base64.StdEncoding.EncodeToString([]byte("anyuser:wrong-key"))
+		req.Header.Set("Authorization", "Basic "+credentials)
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusUnauthorized, w.Code)
+		assert.Contains(t, w.Body.String(), "unauthorized")
+	})
+
+	t.Run("x-api-key and Basic Auth with matching keys", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		req.Header.Set("x-api-key", "valid-key-1")
+		credentials := base64.StdEncoding.EncodeToString([]byte("user:valid-key-1"))
+		req.Header.Set("Authorization", "Basic "+credentials)
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+
+	t.Run("401 response includes WWW-Authenticate header", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusUnauthorized, w.Code)
+		assert.Equal(t, `Basic realm="llama-swap"`, w.Header().Get("WWW-Authenticate"))
+	})
+}
+
+func TestProxyManager_APIKeyAuth_Disabled(t *testing.T) {
+	// Config without RequiredAPIKeys - auth should be disabled
+	testConfig := config.AddDefaultGroupToConfig(config.Config{
+		HealthCheckTimeout: 15,
+		Models: map[string]config.ModelConfig{
+			"model1": getTestSimpleResponderConfig("model1"),
+		},
+		LogLevel: "error",
+	})
+
+	proxy := New(testConfig)
+	defer proxy.StopProcesses(StopImmediately)
+
+	t.Run("requests pass without API key when not configured", func(t *testing.T) {
+		reqBody := `{"model":"model1"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+	})
+}
+
+// TestProxyManager_PeerProxy_InferenceHandler tests the peerProxy integration
+// in proxyInferenceHandler for issue #433
+func TestProxyManager_PeerProxy_InferenceHandler(t *testing.T) {
+	t.Run("requests to peer models are proxied", func(t *testing.T) {
+		// Create a test server to act as the peer
+		peerServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.Header().Set("Content-Type", "application/json")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte(`{"response":"from-peer","model":"peer-model"}`))
+		}))
+		defer peerServer.Close()
+
+		// Create config with peers but no local model for "peer-model"
+		configStr := fmt.Sprintf(`
+logLevel: error
+peers:
+  test-peer:
+    proxy: %s
+    models:
+      - peer-model
+models:
+  local-model:
+    cmd: %s -port ${PORT} -silent -respond local-model
+`, peerServer.URL, getSimpleResponderPath())
+
+		testConfig, err := config.LoadConfigFromReader(strings.NewReader(configStr))
+		assert.NoError(t, err)
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		reqBody := `{"model":"peer-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+		assert.Contains(t, w.Body.String(), "from-peer")
+	})
+
+	t.Run("local models take precedence over peer models", func(t *testing.T) {
+		// Create a test server to act as the peer - should NOT be called
+		peerCalled := false
+		peerServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			peerCalled = true
+			w.Header().Set("Content-Type", "application/json")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte(`{"response":"from-peer"}`))
+		}))
+		defer peerServer.Close()
+
+		// Create config where "shared-model" exists both locally and on peer
+		configStr := fmt.Sprintf(`
+logLevel: error
+peers:
+  test-peer:
+    proxy: %s
+    models:
+      - shared-model
+models:
+  shared-model:
+    cmd: %s -port ${PORT} -silent -respond local-response
+`, peerServer.URL, getSimpleResponderPath())
+
+		testConfig, err := config.LoadConfigFromReader(strings.NewReader(configStr))
+		assert.NoError(t, err)
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		reqBody := `{"model":"shared-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+		assert.Contains(t, w.Body.String(), "local-response")
+		assert.False(t, peerCalled, "peer should not be called when local model exists")
+	})
+
+	t.Run("unknown model returns error", func(t *testing.T) {
+		// Create a test server to act as the peer
+		peerServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusOK)
+		}))
+		defer peerServer.Close()
+
+		configStr := fmt.Sprintf(`
+logLevel: error
+peers:
+  test-peer:
+    proxy: %s
+    models:
+      - peer-model
+models:
+  local-model:
+    cmd: %s -port ${PORT} -silent -respond local-model
+`, peerServer.URL, getSimpleResponderPath())
+
+		testConfig, err := config.LoadConfigFromReader(strings.NewReader(configStr))
+		assert.NoError(t, err)
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		reqBody := `{"model":"unknown-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusBadRequest, w.Code)
+		assert.Contains(t, w.Body.String(), "could not find suitable inference handler")
+	})
+
+	t.Run("peer API key is injected into request", func(t *testing.T) {
+		var receivedAuthHeader string
+		peerServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			receivedAuthHeader = r.Header.Get("Authorization")
+			w.Header().Set("Content-Type", "application/json")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte(`{"response":"ok"}`))
+		}))
+		defer peerServer.Close()
+
+		configStr := fmt.Sprintf(`
+logLevel: error
+peers:
+  test-peer:
+    proxy: %s
+    apiKey: secret-peer-key
+    models:
+      - peer-model
+models:
+  local-model:
+    cmd: %s -port ${PORT} -silent -respond local-model
+`, peerServer.URL, getSimpleResponderPath())
+
+		testConfig, err := config.LoadConfigFromReader(strings.NewReader(configStr))
+		assert.NoError(t, err)
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		reqBody := `{"model":"peer-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+		assert.Equal(t, "Bearer secret-peer-key", receivedAuthHeader)
+	})
+
+	t.Run("no peers configured - unknown model returns error", func(t *testing.T) {
+		testConfig := config.AddDefaultGroupToConfig(config.Config{
+			HealthCheckTimeout: 15,
+			Models: map[string]config.ModelConfig{
+				"local-model": getTestSimpleResponderConfig("local-model"),
+			},
+			LogLevel: "error",
+		})
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		// peerProxy exists but has no peer models configured
+		assert.False(t, proxy.peerProxy.HasPeerModel("unknown-model"))
+
+		reqBody := `{"model":"unknown-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusBadRequest, w.Code)
+		assert.Contains(t, w.Body.String(), "could not find suitable inference handler")
+	})
+
+	t.Run("peer streaming response sets X-Accel-Buffering header", func(t *testing.T) {
+		peerServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.Header().Set("Content-Type", "text/event-stream")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte("data: test\n\n"))
+		}))
+		defer peerServer.Close()
+
+		configStr := fmt.Sprintf(`
+logLevel: error
+peers:
+  test-peer:
+    proxy: %s
+    models:
+      - peer-model
+models:
+  local-model:
+    cmd: %s -port ${PORT} -silent -respond local-model
+`, peerServer.URL, getSimpleResponderPath())
+
+		testConfig, err := config.LoadConfigFromReader(strings.NewReader(configStr))
+		assert.NoError(t, err)
+
+		proxy := New(testConfig)
+		defer proxy.StopProcesses(StopImmediately)
+
+		reqBody := `{"model":"peer-model"}`
+		req := httptest.NewRequest("POST", "/v1/chat/completions", bytes.NewBufferString(reqBody))
+		w := CreateTestResponseRecorder()
+
+		proxy.ServeHTTP(w, req)
+		assert.Equal(t, http.StatusOK, w.Code)
+		assert.Equal(t, "no", w.Header().Get("X-Accel-Buffering"))
+	})
+}

ui/src/contexts/APIProvider.tsx

@@ -10,6 +10,7 @@ export interface Model {
   name: string;
   description: string;
   unlisted: boolean;
+  peerID: string;
 }
 
 interface APIProviderType {
@@ -70,7 +71,7 @@ export function APIProvider({ children, autoStartAPIEvents = true }: APIProvider
   const [versionInfo, setVersionInfo] = useState<VersionInfo>({
     build_date: "unknown",
     commit: "unknown",
-    version: "unknown"
+    version: "unknown",
   });
   //const apiEventSource = useRef<EventSource | null>(null);
 
@@ -166,7 +167,7 @@ export function APIProvider({ children, autoStartAPIEvents = true }: APIProvider
   }, []);
 
   useEffect(() => {
-    // fetch version 
+    // fetch version
     const fetchVersion = async () => {
       try {
         const response = await fetch("/api/version");
@@ -180,7 +181,7 @@ export function APIProvider({ children, autoStartAPIEvents = true }: APIProvider
       }
     };
 
-    if (connectionStatus === 'connected') {
+    if (connectionStatus === "connected") {
       fetchVersion();
     }
   }, [connectionStatus]);
@@ -265,7 +266,19 @@ export function APIProvider({ children, autoStartAPIEvents = true }: APIProvider
       connectionStatus,
       versionInfo,
     }),
-    [models, listModels, unloadAllModels, unloadSingleModel, loadModel, enableAPIEvents, proxyLogs, upstreamLogs, metrics, connectionStatus, versionInfo]
+    [
+      models,
+      listModels,
+      unloadAllModels,
+      unloadSingleModel,
+      loadModel,
+      enableAPIEvents,
+      proxyLogs,
+      upstreamLogs,
+      metrics,
+      connectionStatus,
+      versionInfo,
+    ]
   );
 
   return <APIContext.Provider value={value}>{children}</APIContext.Provider>;

ui/src/pages/Models.tsx

@@ -44,8 +44,24 @@ function ModelsPanel() {
   const [showIdorName, setShowIdorName] = usePersistentState<"id" | "name">("showIdorName", "id"); // true = show ID, false = show name
   const [menuOpen, setMenuOpen] = useState(false);
 
-  const filteredModels = useMemo(() => {
-    return models.filter((model) => showUnlisted || !model.unlisted);
+  const { regularModels, peerModelsByPeerId } = useMemo(() => {
+    const filtered = models.filter((model) => showUnlisted || !model.unlisted);
+    const peerModels = filtered.filter((m) => m.peerID);
+
+    // Group peer models by peerID
+    const grouped = peerModels.reduce((acc, model) => {
+      const peerId = model.peerID || "unknown";
+      if (!acc[peerId]) {
+        acc[peerId] = [];
+      }
+      acc[peerId].push(model);
+      return acc;
+    }, {} as Record<string, typeof peerModels>);
+
+    return {
+      regularModels: filtered.filter((m) => !m.peerID),
+      peerModelsByPeerId: grouped,
+    };
   }, [models, showUnlisted]);
 
   const handleUnloadAllModels = useCallback(async () => {
@@ -151,7 +167,7 @@ function ModelsPanel() {
             </tr>
           </thead>
           <tbody>
-            {filteredModels.map((model) => (
+            {regularModels.map((model) => (
               <tr key={model.id} className="border-b hover:bg-secondary-hover border-gray-200">
                 <td className={`${model.unlisted ? "text-txtsecondary" : ""}`}>
                   <a href={`/upstream/${model.id}/`} className="font-semibold" target="_blank">
@@ -186,6 +202,34 @@ function ModelsPanel() {
             ))}
           </tbody>
         </table>
+
+        {Object.keys(peerModelsByPeerId).length > 0 && (
+          <>
+            <h3 className="mt-8 mb-2">Peer Models</h3>
+            {Object.entries(peerModelsByPeerId)
+              .sort(([a], [b]) => a.localeCompare(b))
+              .map(([peerId, models]) => (
+                <div key={peerId} className="mb-4">
+                  <table className="w-full">
+                    <thead className="sticky top-0 bg-card z-10">
+                      <tr className="text-left border-b border-gray-200 dark:border-white/10 bg-surface">
+                        <th className="font-semibold">{peerId}</th>
+                      </tr>
+                    </thead>
+                    <tbody>
+                      {models.map((model) => (
+                        <tr key={model.id} className="border-b hover:bg-secondary-hover border-gray-200">
+                          <td className={`pl-8 ${model.unlisted ? "text-txtsecondary" : ""}`}>
+                            <span>{model.id}</span>
+                          </td>
+                        </tr>
+                      ))}
+                    </tbody>
+                  </table>
+                </div>
+              ))}
+          </>
+        )}
       </div>
     </div>
   );
@@ -223,11 +267,7 @@ function TokenHistogram({ data }: { data: HistogramData }) {
 
   return (
     <div className="mt-2 w-full">
-      <svg
-        viewBox={`0 0 ${viewBoxWidth} ${height}`}
-        className="w-full h-auto"
-        preserveAspectRatio="xMidYMid meet"
-      >
+      <svg viewBox={`0 0 ${viewBoxWidth} ${height}`} className="w-full h-auto" preserveAspectRatio="xMidYMid meet">
         {/* Y-axis */}
         <line
           x1={padding.left}
@@ -312,14 +352,7 @@ function TokenHistogram({ data }: { data: HistogramData }) {
         />
 
         {/* X-axis labels */}
-        <text
-          x={padding.left}
-          y={height - 5}
-          fontSize="10"
-          fill="currentColor"
-          opacity="0.6"
-          textAnchor="start"
-        >
+        <text x={padding.left} y={height - 5} fontSize="10" fill="currentColor" opacity="0.6" textAnchor="start">
           {min.toFixed(1)}
         </text>