refactor(config): fix LLM timeout architecture

[cpcloud]

Summary

Closes #732.

• Clarify timeout semantics: ResolvedLLM.Timeout is now the inference context deadline (was the HTTP client timeout). Per-pipeline overrides (llm.chat.timeout, llm.extraction.timeout) inherit from the base llm.timeout.
• Derive HTTP client timeout: NewClient computes max(timeout, QuickOpTimeout) so quick ops (ping, model listing) aren't killed by short inference timeouts. QuickOpTimeout stays as a non-configurable 30s constant.
• Add chat inference deadline: Chat streaming now uses context.WithTimeout (was WithCancel with no deadline), enforcing the configured llm.chat.timeout.
• Deprecate extraction.llm_timeout: Migrated to llm.extraction.timeout with TOML key migration and env var rename (MICASA_EXTRACTION_LLM_TIMEOUT -> MICASA_LLM_EXTRACTION_TIMEOUT).
• Remove redundant LLMInferenceTimeout: extractionConfig.Timeout now serves as the inference deadline directly, eliminating the extra field.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 61.22449% with 19 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.61%. Comparing base (847153e) to head (0422e3d).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
internal/app/types.go	0.00%	10 Missing ⚠️
internal/app/chat.go	30.00%	6 Missing and 1 partial ⚠️
internal/app/extraction.go	50.00%	2 Missing ⚠️

Additional details and impacted files

Files with missing lines	Coverage Δ
cmd/micasa/main.go	3.33% <ø> (+0.01%)	⬆️
internal/app/model.go	62.36% <100.00%> (-0.02%)	⬇️
internal/config/config.go	88.02% <100.00%> (-1.62%)	⬇️
internal/config/show.go	81.69% <ø> (ø)
internal/llm/client.go	91.40% <100.00%> (+0.05%)	⬆️
internal/app/extraction.go	70.71% <50.00%> (-0.05%)	⬇️
internal/app/chat.go	72.34% <30.00%> (-0.13%)	⬇️
internal/app/types.go	52.94% <0.00%> (+0.61%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR refactors the LLM timeout architecture (issue #732) to cleanly separate two concerns: quick-operation timeouts (ping, model listing) and per-pipeline inference timeouts (chat streaming, extraction). Previously, llm.timeout served double duty as both an HTTP client timeout and an inference deadline, which conflated two different concerns and created confusing interactions.

Changes:

• Clarified ResolvedLLM.Timeout semantics: now represents the inference context deadline; HTTP client timeout is derived as max(timeout, QuickOpTimeout) in NewClient
• Added chat inference deadline: chat streaming now uses context.WithTimeout instead of context.WithCancel, enforcing the configured llm.chat.timeout
• Deprecated extraction.llm_timeout: migrated to llm.extraction.timeout with TOML key migration, env var rename, and documentation update; removed the now-redundant LLMInferenceTimeout field

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
plans/llm-timeout-architecture.md	New planning document describing the problem, design, and changes for the timeout refactor
internal/llm/client.go	NewClient now derives HTTP client timeout as max(timeout, QuickOpTimeout); updated doc comment and timeout error message
internal/config/config.go	Clarified LLM.Timeout and ResolvedLLM.Timeout doc comments; added extraction.llm_timeout TOML migration and MICASA_EXTRACTION_LLM_TIMEOUT env var rename entry
internal/app/types.go	Removed LLMInferenceTimeout from extractionConfig; Timeout field (now annotated as inference context deadline) absorbs its role; updated SetExtraction signature
internal/app/model.go	Removed llmInferenceTimeout initialization (field no longer exists)
internal/app/extraction.go	llmExtractCmd now reads extractionTimeout directly; llmPingCmd uses client.Timeout() for the quick-op deadline
internal/app/chat.go	All three chat stream paths now use context.WithTimeout via new chatInferenceTimeout() helper
internal/app/extraction_test.go	Updated test assertion from old extraction.llm_timeout key name to new llm.extraction.timeout
cmd/micasa/main.go	Removed the now-redundant cfg.Extraction.LLMTimeoutDuration() argument from SetExtraction call
docs/content/docs/reference/configuration.md	Updated all references from old to new config keys; added deprecated env vars table entry

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.