feat(client): add LLMClient abstraction with configurable reasoning inclusion #52

Merged

ph merged 1 commit from feat/issue-49-llm-client into master

2026-06-08 15:30:11 +00:00

ph commented

2026-06-08 15:29:42 +00:00

Owner

Summary

Add an LLMClient wrapper over go-openai that centralizes all LLM API calls and supports configurable reasoning content filtering.

Closes #49.

Changes

File	Change
`internal/config/config.go`	Add `IncludeReasoning bool` config field (default false)
`internal/client/llm_client.go`	New — `LLMClient` + `ChatRequest` + `convertMessages()`
`internal/client/llm_client_test.go`	New — 7 unit tests
`internal/client/manager.go`	Add `LLMClient()` factory method
`internal/agent/agent.go`	Switch from `openai.Client` to `LLMClient`
`internal/agent/compact.go`	Switch to `LLMClient`

Design

All LLM calls (streaming and non-streaming) now go through LLMClient
Reasoning content is not sent to the API by default (IncludeReasoning = false)
Configurable via ~/.config/gd/config.toml: [defaults] include_reasoning = true
Stream capture and frontend rendering of thinking parts are unaffected

Verification

make build ✅
go test ./... -count=1 — all 8 packages pass ✅
go vet ✅

## Summary Add an `LLMClient` wrapper over `go-openai` that centralizes all LLM API calls and supports configurable reasoning content filtering. Closes #49. ## Changes | File | Change | |------|--------| | `internal/config/config.go` | Add `IncludeReasoning bool` config field (default false) | | `internal/client/llm_client.go` | **New** — `LLMClient` + `ChatRequest` + `convertMessages()` | | `internal/client/llm_client_test.go` | **New** — 7 unit tests | | `internal/client/manager.go` | Add `LLMClient()` factory method | | `internal/agent/agent.go` | Switch from `*openai.Client` to `*LLMClient` | | `internal/agent/compact.go` | Switch to `LLMClient` | ## Design - All LLM calls (streaming and non-streaming) now go through `LLMClient` - Reasoning content is **not** sent to the API by default (`IncludeReasoning = false`) - Configurable via `~/.config/gd/config.toml`: `[defaults] include_reasoning = true` - Stream capture and frontend rendering of thinking parts are unaffected ## Verification - `make build` ✅ - `go test ./... -count=1` — all 8 packages pass ✅ - `go vet` ✅

ph added 1 commit

2026-06-08 15:29:42 +00:00

feat(client): add LLMClient abstraction with configurable reasoning inclusion 65a4aae79d

Add LLMClient wrapper over go-openai that centralizes all LLM API
calls and supports configurable reasoning content filtering.

- Add IncludeReasoning bool to config (default false)
- Add LLMClient + ChatRequest + convertMessages() in internal/client
- Add LLMClient() factory method on Manager
- Switch Agent from openai.Client to LLMClient
- Add 7 unit tests for message conversion with/without reasoning
- Remove dead code: Agent.modelID(), ModelTypeToModel()

Closes #49

ph merged commit b94cdaac57 into master

2026-06-08 15:30:11 +00:00

ph deleted branch feat/issue-49-llm-client

2026-06-08 15:30:11 +00:00

ph referenced this pull request from a commit

2026-06-08 15:30:13 +00:00

Merge pull request 'feat(client): add LLMClient abstraction with configurable reasoning inclusion' (#52) from feat/issue-49-llm-client into master