Model capabilities 734 (#842)

internal/config,server: implement model capabilities - define the capabilities of a model using a simple config block on the model - v1/models renders out capabilities to be compatible with openrouter, huggingface chat, and mistral formats for broader compatibility - add support for capabilities in UI Fixes #734
2026-06-13 23:23:19 -07:00
parent 62aea0e83d
commit 92b90447e8
16 changed files with 868 additions and 35 deletions
@@ -447,6 +447,10 @@ func LoadConfigFromReader(r io.Reader) (Config, error) {
 			}
 		}

+		if err = modelConfig.Capabilities.Validate(); err != nil {
+			return Config{}, fmt.Errorf("model %s: %w", modelId, err)
+		}
+
 		// Validate SetParamsByID keys and values
 		for key, paramMap := range modelConfig.Filters.SetParamsByID {
 			if matches := macroPatternRegex.FindAllStringSubmatch(key, -1); len(matches) > 0 {
@@ -2,6 +2,7 @@ package config

 import (
 	"errors"
+	"fmt"
 	"runtime"
 )

@@ -9,6 +10,47 @@ const (
 	MODEL_CONFIG_DEFAULT_TTL = -1
 )

+var validModalities = map[string]struct{}{
+	"text":  {},
+	"audio": {},
+	"image": {},
+}
+
+// ModelCapConfig defines what modalities and features a model supports.
+// Used in /v1/models to inform clients. An empty block (all zero values) is
+// treated as not configured.
+type ModelCapConfig struct {
+	In       []string `yaml:"in"`
+	Out      []string `yaml:"out"`
+	Tools    bool     `yaml:"tools"`
+	Reranker bool     `yaml:"reranker"`
+	Context  int      `yaml:"context"`
+}
+
+// Empty returns true when all fields are at their zero values.
+func (c ModelCapConfig) Empty() bool {
+	return len(c.In) == 0 && len(c.Out) == 0 && !c.Tools && !c.Reranker && c.Context == 0
+}
+
+// Validate checks that all modality values are recognized and context is
+// non-negative. Returns an error if any value is invalid.
+func (c ModelCapConfig) Validate() error {
+	for _, m := range c.In {
+		if _, ok := validModalities[m]; !ok {
+			return fmt.Errorf("capabilities.in: invalid modality %q, must be one of: text, audio, image", m)
+		}
+	}
+	for _, m := range c.Out {
+		if _, ok := validModalities[m]; !ok {
+			return fmt.Errorf("capabilities.out: invalid modality %q, must be one of: text, audio, image", m)
+		}
+	}
+	if c.Context < 0 {
+		return errors.New("capabilities.context: must be >= 0")
+	}
+	return nil
+}
+
 // TimeoutsConfig holds timeout settings for proxy connections
 // 0 = no timeout
 type TimeoutsConfig struct {
@@ -55,6 +97,9 @@ type ModelConfig struct {
 	// Timeout settings for proxy connections
 	Timeouts TimeoutsConfig `yaml:"timeouts"`

+	// Capabilities defines what modalities and features the model supports.
+	Capabilities ModelCapConfig `yaml:"capabilities"`
+
 	// Copy of HealthCheckTimeout from global config
 	HealthCheckTimeout int `yaml:"healthCheckTimeout"`
 }
@@ -152,7 +152,7 @@ models:
        stop:
          - "<|end|>"
          - "<|stop|>"
-`
+ `
 	config, err := LoadConfigFromReader(strings.NewReader(content))
 	assert.NoError(t, err)

@@ -170,3 +170,167 @@ models:
 	assert.Equal(t, 0.7, setParams["temperature"])
 	assert.Equal(t, 0.9, setParams["top_p"])
 }
+
+func TestConfig_ModelCapabilities(t *testing.T) {
+	t.Run("all fields", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      in:
+        - text
+        - audio
+        - image
+      out:
+        - text
+        - audio
+        - image
+      tools: true
+      context: 32000
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.False(t, mc.Capabilities.Empty())
+		assert.Equal(t, []string{"text", "audio", "image"}, mc.Capabilities.In)
+		assert.Equal(t, []string{"text", "audio", "image"}, mc.Capabilities.Out)
+		assert.True(t, mc.Capabilities.Tools)
+		assert.Equal(t, 32000, mc.Capabilities.Context)
+	})
+
+	t.Run("partial fields", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      tools: true
+      context: 8192
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.False(t, mc.Capabilities.Empty())
+		assert.Nil(t, mc.Capabilities.In)
+		assert.Nil(t, mc.Capabilities.Out)
+		assert.True(t, mc.Capabilities.Tools)
+		assert.Equal(t, 8192, mc.Capabilities.Context)
+	})
+
+	t.Run("not set", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.True(t, mc.Capabilities.Empty())
+	})
+
+	t.Run("tools false is empty", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      tools: false
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.True(t, mc.Capabilities.Empty())
+	})
+
+	t.Run("reranker true is not empty", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      reranker: true
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.False(t, mc.Capabilities.Empty())
+		assert.True(t, mc.Capabilities.Reranker)
+	})
+
+	t.Run("reranker false is empty", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      reranker: false
+`
+		config, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.NoError(t, err)
+
+		mc := config.Models["model1"]
+		assert.True(t, mc.Capabilities.Empty())
+	})
+}
+
+func TestConfig_ModelCapabilities_Validate(t *testing.T) {
+	t.Run("valid_modalities", func(t *testing.T) {
+		caps := ModelCapConfig{
+			In:      []string{"text", "image"},
+			Out:     []string{"text", "audio"},
+			Tools:   true,
+			Context: 100000,
+		}
+		assert.NoError(t, caps.Validate())
+	})
+
+	t.Run("empty_is_valid", func(t *testing.T) {
+		caps := ModelCapConfig{}
+		assert.NoError(t, caps.Validate())
+	})
+
+	t.Run("invalid_in_modality", func(t *testing.T) {
+		caps := ModelCapConfig{In: []string{"video"}}
+		err := caps.Validate()
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "capabilities.in")
+		assert.Contains(t, err.Error(), "video")
+	})
+
+	t.Run("invalid_out_modality", func(t *testing.T) {
+		caps := ModelCapConfig{Out: []string{"video"}}
+		err := caps.Validate()
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "capabilities.out")
+		assert.Contains(t, err.Error(), "video")
+	})
+
+	t.Run("negative_context", func(t *testing.T) {
+		caps := ModelCapConfig{Context: -1}
+		err := caps.Validate()
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "capabilities.context")
+	})
+
+	t.Run("rejects_invalid_at_load", func(t *testing.T) {
+		content := `
+models:
+  model1:
+    cmd: path/to/cmd --port ${PORT}
+    capabilities:
+      in:
+        - text
+        - video
+`
+		_, err := LoadConfigFromReader(strings.NewReader(content))
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "video")
+	})
+}