feat: live-validated against Ollama Cloud; schema instruction fallback for cloud

Phase 8: all six live checks pass (tier aliases, thinking-tier chat, real
tool invocation, structured Generate[T], forced failover with bench+skip,
skill agent). Discovery: ollama.com ignores the format field — the
provider now also states the schema as a system instruction (constrained
decoding locally, instruction-guided JSON on cloud), with hermetic test.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-06-10 13:22:54 +02:00
parent 97513141dc
commit 04b21fdad2
5 changed files with 299 additions and 1 deletions
+5 -1
View File
@@ -281,7 +281,11 @@ streaming); majordomo's stream API works against it and delivers the
response as a single delta plus final event.
Notes: Ollama has no native tool_choice — `"none"` drops the tools;
`"required"`/named choices are best-effort ignored there.
`"required"`/named choices are best-effort ignored there. Ollama Cloud
ignores the `format` field (verified live), so the provider also states
the schema as an explicit system instruction — constrained decoding on
local Ollama, instruction-guided JSON on cloud, one canonical API either
way.
Cross-cutting: Parse grammar ✅ · aliases/tiers ✅ · failover chains ✅ ·
health tracking/backoff ✅ · LLM_* env DSNs ✅ · media pipeline ✅