mirror of
https://github.com/tiennm99/goclaw.git
synced 2026-10-11 12:18:59 +00:00
Duy /zuey/ authored and GitHub committed
1 parent
85fe1b8e81
commit
4f87b93dbd
16 files changed
+451
No files matched your search
@@ -91,6 +91,17 @@ ui/desktop/ Wails v2 desktop app (React frontend + embedded ga
|
||||
- **Telegram formatting:** LLM output → `SanitizeAssistantContent()` → `markdownToTelegramHTML()` → `chunkHTML()` → `sendHTML()`. Tables rendered as ASCII in `<pre>` tags
|
||||
- **i18n:** Web UI uses `i18next` with namespace-split locale files in `ui/web/src/i18n/locales/{lang}/`. Backend uses `internal/i18n` message catalog with `i18n.T(locale, key, args...)`. Locale propagated via `store.WithLocale(ctx)` — WS `connect` param `locale`, HTTP `Accept-Language` header. Supported: en (default), vi, zh. New user-facing strings: add key to `internal/i18n/keys.go`, add translations to all 3 catalog files. New UI strings: add key to all 3 locale dirs. Bootstrap templates (SOUL.md, etc.) stay English-only (LLM consumption).
|
||||
|
||||
## Cross-Surface Feature/Fix Parity
|
||||
|
||||
Every feature implementation or bug fix must audit and update all affected product surfaces before it is marked complete:
|
||||
|
||||
- **Gateway server:** handlers, WebSocket methods, stores, migrations, provider/runtime behavior, background jobs.
|
||||
- **API contract:** request/response structs, `pkg/protocol`, OpenAPI/docs, compatibility shims, tests built from real response shapes.
|
||||
- **Web UI:** `ui/web` screens, hooks, i18n, validation, loading/error states, and contract consumers.
|
||||
- **CLI/runtime package:** `cmd` commands, operator/runtime package commands, installers/manifests, response parsers, and runtime package docs.
|
||||
|
||||
Do not ship a backend-only change when the web UI, CLI/runtime package, or API contract must also change. If a surface is not affected, state `Surface parity: <surface> N/A because ...` in the plan, PR, or final report. For cross-repo CLI work, verify the current CLI/runtime package repo and release channel before claiming parity.
|
||||
|
||||
## Running
|
||||
|
||||
```bash
|
||||
|
||||
@@ -91,6 +91,17 @@ ui/desktop/ Wails v2 desktop app (React frontend + embedded ga
|
||||
- **Telegram formatting:** LLM output → `SanitizeAssistantContent()` → `markdownToTelegramHTML()` → `chunkHTML()` → `sendHTML()`. Tables rendered as ASCII in `<pre>` tags
|
||||
- **i18n:** Web UI uses `i18next` with namespace-split locale files in `ui/web/src/i18n/locales/{lang}/`. Backend uses `internal/i18n` message catalog with `i18n.T(locale, key, args...)`. Locale propagated via `store.WithLocale(ctx)` — WS `connect` param `locale`, HTTP `Accept-Language` header. Supported: en (default), vi, zh. New user-facing strings: add key to `internal/i18n/keys.go`, add translations to all 3 catalog files. New UI strings: add key to all 3 locale dirs. Bootstrap templates (SOUL.md, etc.) stay English-only (LLM consumption).
|
||||
|
||||
## Cross-Surface Feature/Fix Parity
|
||||
|
||||
Every feature implementation or bug fix must audit and update all affected product surfaces before it is marked complete:
|
||||
|
||||
- **Gateway server:** handlers, WebSocket methods, stores, migrations, provider/runtime behavior, background jobs.
|
||||
- **API contract:** request/response structs, `pkg/protocol`, OpenAPI/docs, compatibility shims, tests built from real response shapes.
|
||||
- **Web UI:** `ui/web` screens, hooks, i18n, validation, loading/error states, and contract consumers.
|
||||
- **CLI/runtime package:** `cmd` commands, operator/runtime package commands, installers/manifests, response parsers, and runtime package docs.
|
||||
|
||||
Do not ship a backend-only change when the web UI, CLI/runtime package, or API contract must also change. If a surface is not affected, state `Surface parity: <surface> N/A because ...` in the plan, PR, or final report. For cross-repo CLI work, verify the current CLI/runtime package repo and release channel before claiming parity.
|
||||
|
||||
## Running
|
||||
|
||||
```bash
|
||||
|
||||
@@ -351,6 +351,11 @@ Standard OpenAI-compatible provider targeting the Alibaba Coding API.
|
||||
|
||||
- **Default model**: `qwen3.5-plus`
|
||||
- **Base URL**: `https://coding-intl.dashscope.aliyuncs.com/v1`
|
||||
- **Catalog source**: hardcoded because the Coding API does not expose a standard `/v1/models` endpoint
|
||||
|
||||
| Model | Display name | Capabilities |
|
||||
|-------|--------------|--------------|
|
||||
| `qwen3.7-plus` | Qwen 3.7 Plus | Text Generation, Deep Thinking, Visual Understanding |
|
||||
|
||||
---
|
||||
|
||||
|
||||
@@ -139,6 +139,12 @@ Enables thinking via `enable_thinking: true` plus a `thinking_budget` parameter.
|
||||
|
||||
Other models (e.g., `qwen3-plus`, `qwen3-turbo`) silently skip thinking injection to avoid API errors.
|
||||
|
||||
**Bailian Coding note**: Bailian is a separate OpenAI-compatible Coding
|
||||
endpoint. Its model catalog includes `qwen3.7-plus` with Deep Thinking and
|
||||
Visual Understanding listed for selection, but this DashScope `enable_thinking`
|
||||
/ `thinking_budget` injection path does not apply to Bailian unless the Coding
|
||||
endpoint's explicit request controls are verified separately.
|
||||
|
||||
**Important limitation**: DashScope does not support streaming when tools are present. When an agent has tools enabled and thinking is active, the provider automatically falls back to non-streaming mode (single `Chat()` call) and synthesizes chunk callbacks to maintain the event flow.
|
||||
|
||||
### Codex (ChatGPT OAuth Responses API)
|
||||
|
||||
@@ -6,6 +6,21 @@ Significant changes, features, and fixes in reverse chronological order.
|
||||
|
||||
## 2026-06-12
|
||||
|
||||
### Bailian Coding qwen3.7-plus catalog (issue #169)
|
||||
|
||||
**Changes**
|
||||
|
||||
- Added `qwen3.7-plus` / `Qwen 3.7 Plus` to the hardcoded Bailian
|
||||
Coding provider model catalog.
|
||||
- Documented the model's advertised Text Generation, Deep Thinking, and Visual
|
||||
Understanding capabilities while keeping Bailian on the existing
|
||||
OpenAI-compatible provider wrapper.
|
||||
|
||||
**Tests**
|
||||
|
||||
- Added provider model endpoint coverage to verify Bailian exposes
|
||||
`qwen3.7-plus` without adding unsupported OpenAI reasoning metadata.
|
||||
|
||||
### Multi-attachment outbound delivery (issue #172)
|
||||
|
||||
**Changes**
|
||||
|
||||
@@ -7,6 +7,8 @@ import "github.com/nextlevelbuilder/goclaw/internal/providers"
|
||||
// The platform does not expose a /v1/models endpoint.
|
||||
func bailianModels() []ModelInfo {
|
||||
return []ModelInfo{
|
||||
// qwen3.7-plus: Text Generation + Deep Thinking + Visual Understanding.
|
||||
{ID: "qwen3.7-plus", Name: "Qwen 3.7 Plus"},
|
||||
{ID: "qwen3.6-plus", Name: "Qwen 3.6 Plus"},
|
||||
{ID: "qwen3.5-plus", Name: "Qwen 3.5 Plus"},
|
||||
{ID: "kimi-k2.5", Name: "Kimi K2.5"},
|
||||
|
||||
@@ -116,6 +116,54 @@ func TestProvidersHandlerListProviderModelsChatGPTOAuthIncludesReasoningMetadata
|
||||
}
|
||||
}
|
||||
|
||||
func TestProvidersHandlerListProviderModelsBailianIncludesQwen37Plus(t *testing.T) {
|
||||
token := setupProvidersAdminToken(t)
|
||||
providerStore := newMockProviderStore()
|
||||
provider := &store.LLMProviderData{
|
||||
BaseModel: store.BaseModel{ID: uuid.New()},
|
||||
Name: "bailian-coding",
|
||||
ProviderType: store.ProviderBailian,
|
||||
APIKey: "token",
|
||||
Enabled: true,
|
||||
}
|
||||
if err := providerStore.CreateProvider(t.Context(), provider); err != nil {
|
||||
t.Fatalf("CreateProvider() error = %v", err)
|
||||
}
|
||||
|
||||
handler := NewProvidersHandler(providerStore, newMockSecretsStore(), nil, "")
|
||||
mux := http.NewServeMux()
|
||||
handler.RegisterRoutes(mux)
|
||||
|
||||
req := httptest.NewRequest(http.MethodGet, "/v1/providers/"+provider.ID.String()+"/models", nil)
|
||||
req.Header.Set("Authorization", "Bearer "+token)
|
||||
w := httptest.NewRecorder()
|
||||
mux.ServeHTTP(w, req)
|
||||
|
||||
if w.Code != http.StatusOK {
|
||||
t.Fatalf("status code = %d, want %d, body=%s", w.Code, http.StatusOK, w.Body.String())
|
||||
}
|
||||
|
||||
var result ProviderModelsResponse
|
||||
if err := json.NewDecoder(w.Body).Decode(&result); err != nil {
|
||||
t.Fatalf("Decode() error = %v", err)
|
||||
}
|
||||
|
||||
for _, model := range result.Models {
|
||||
if model.ID != "qwen3.7-plus" {
|
||||
continue
|
||||
}
|
||||
if model.Name != "Qwen 3.7 Plus" {
|
||||
t.Fatalf("model name = %q, want Qwen 3.7 Plus", model.Name)
|
||||
}
|
||||
if model.Reasoning != nil {
|
||||
t.Fatalf("model reasoning = %#v, want nil for Bailian OpenAI-compatible catalog entry", model.Reasoning)
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
t.Fatalf("qwen3.7-plus not found in Bailian model list: %#v", result.Models)
|
||||
}
|
||||
|
||||
func TestProvidersHandlerListProviderModelsOpenAICompatAnnotatesKnownModels(t *testing.T) {
|
||||
token := setupProvidersAdminToken(t)
|
||||
upstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
||||
|
||||
@@ -0,0 +1,56 @@
|
||||
---
|
||||
phase: 1
|
||||
title: Catalog and capability docs
|
||||
status: completed
|
||||
priority: P2
|
||||
effort: 30m
|
||||
dependencies: []
|
||||
---
|
||||
|
||||
# Phase 1: Catalog and capability docs
|
||||
|
||||
## Overview
|
||||
|
||||
Add the new Bailian model to the backend catalog and document its advertised capabilities without changing runtime request-body semantics.
|
||||
|
||||
## Requirements
|
||||
|
||||
- Functional: `bailianModels()` returns `{ID: "qwen3.7-plus", Name: "Qwen 3.7 Plus"}`.
|
||||
- Functional: docs state the model capabilities from issue #169: Text Generation, Deep Thinking, Visual Understanding.
|
||||
- Non-functional: no new API schema fields unless already supported by `ModelInfo`.
|
||||
- Non-functional: no runtime behavior change for existing Bailian models.
|
||||
|
||||
## Architecture
|
||||
|
||||
`GET /v1/providers/{id}/models` loads a provider from the store, dispatches `provider_type == "bailian"` to `bailianModels()`, and returns the list. Web and desktop provider model pickers already consume this endpoint, so the backend catalog update is the UI propagation path.
|
||||
|
||||
Do not add `qwen3.7-plus` to `internal/providers/reasoning_capability.go`: `OpenAIProvider.buildRequestBody()` uses that registry to decide whether to send OpenAI `reasoning_effort`. Bailian is registered as `NewOpenAIProvider(...)`, not `NewDashScopeProvider(...)`, so Qwen Deep Thinking docs should not imply a wire-level parameter implementation here.
|
||||
|
||||
## Related Code Files
|
||||
|
||||
- Modify: `internal/http/provider_models_catalog.go`
|
||||
- Modify: `docs/02-providers.md`
|
||||
- Modify: `docs/12-extended-thinking.md`
|
||||
- No UI changes: web and desktop use the existing provider models endpoint.
|
||||
|
||||
## Implementation Steps
|
||||
|
||||
1. Add `qwen3.7-plus` near the current Qwen Plus entries in `bailianModels()`.
|
||||
2. Add concise inline catalog grouping that records Bailian capabilities for Qwen Plus models.
|
||||
3. Update `docs/02-providers.md` Bailian section with the new model and capability row.
|
||||
4. Update `docs/12-extended-thinking.md` to clarify Bailian advertises Deep Thinking for `qwen3.7-plus`, but GoClaw does not inject DashScope `enable_thinking` controls for Bailian in this change.
|
||||
|
||||
## Success Criteria
|
||||
|
||||
- [ ] Catalog contains `qwen3.7-plus`.
|
||||
- [ ] Docs mention Text Generation, Deep Thinking, Visual Understanding.
|
||||
- [ ] No new provider request-body controls are introduced.
|
||||
|
||||
## Risk Assessment
|
||||
|
||||
Risk: adding reasoning metadata could make Bailian receive unsupported OpenAI `reasoning_effort`.
|
||||
Mitigation: leave `LookupReasoningCapability()` unchanged and document the distinction.
|
||||
|
||||
## Security Considerations
|
||||
|
||||
No new secrets, auth paths, or tenant writes.
|
||||
@@ -0,0 +1,47 @@
|
||||
---
|
||||
phase: 2
|
||||
title: Tests and verification
|
||||
status: completed
|
||||
priority: P2
|
||||
effort: 30m
|
||||
dependencies:
|
||||
- 1
|
||||
---
|
||||
|
||||
# Phase 2: Tests and verification
|
||||
|
||||
## Overview
|
||||
|
||||
Add regression coverage for Bailian model discovery and run focused validation for the touched backend/docs paths.
|
||||
|
||||
## Requirements
|
||||
|
||||
- Functional: endpoint-level test proves Bailian returns `qwen3.7-plus`.
|
||||
- Functional: test verifies no unsupported reasoning metadata is exposed for this OpenAI-compatible Bailian model.
|
||||
- Non-functional: targeted Go package tests pass with the installed Go binary.
|
||||
|
||||
## Implementation Steps
|
||||
|
||||
1. Add a provider models test in `internal/http/provider_models_test.go` for `store.ProviderBailian`.
|
||||
2. Use existing mock provider store and auth token helpers.
|
||||
3. Assert HTTP 200, model ID/display name, and nil `Reasoning` for `qwen3.7-plus`.
|
||||
4. Run focused test:
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/http -run 'Bailian|ProviderModels' -count=1`
|
||||
5. Run compile checks when feasible:
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/http -count=1`
|
||||
- broader build/test if the focused package uncovers shared changes.
|
||||
|
||||
## Success Criteria
|
||||
|
||||
- [ ] Focused test fails before catalog change and passes after.
|
||||
- [ ] `go test ./internal/http` passes.
|
||||
- [ ] No generated or unrelated files changed.
|
||||
|
||||
## Risk Assessment
|
||||
|
||||
Risk: test helper requires an API key for Bailian before returning catalog.
|
||||
Mitigation: set a dummy API key on the mock provider because the handler checks non-empty `APIKey` before dispatch.
|
||||
|
||||
## Security Considerations
|
||||
|
||||
Use dummy values only; do not read or write env/secrets.
|
||||
@@ -0,0 +1,46 @@
|
||||
---
|
||||
phase: 3
|
||||
title: "Ship beta PR"
|
||||
status: pending
|
||||
priority: P2
|
||||
effort: "30m"
|
||||
dependencies: [2]
|
||||
---
|
||||
|
||||
# Phase 3: Ship beta PR
|
||||
|
||||
## Overview
|
||||
|
||||
Ship the completed issue #169 work as a beta-targeted PR to `dev`, with labels updated according to `ck:vibe --beta`.
|
||||
|
||||
## Requirements
|
||||
|
||||
- Functional: GitHub issue gets a plan/update comment and `ready to cook` before implementation.
|
||||
- Functional: PR targets `dev` and references issue #169.
|
||||
- Functional: after local and PR review gates pass, source issue and PR get `ready to ship beta`.
|
||||
- Non-functional: no merge is performed because the user invoked `--beta`, not `--ship --beta`.
|
||||
|
||||
## Implementation Steps
|
||||
|
||||
1. Ensure GitHub labels exist: `ready to cook`, `ready to ship stable`, `ready to ship beta`.
|
||||
2. Comment on issue #169 with branch, relative plan path, beta mode, and acceptance criteria.
|
||||
3. After implementation/tests/review, commit with a focused conventional message.
|
||||
4. Push `codex/issue-169-qwen37-bailian-model`.
|
||||
5. Create beta PR against `dev`.
|
||||
6. Review/fix/reply PR feedback and wait for terminal checks when available.
|
||||
7. Add `ready to ship beta` to both issue and PR; remove `ready to cook`.
|
||||
|
||||
## Success Criteria
|
||||
|
||||
- [ ] PR exists and targets `dev`.
|
||||
- [ ] PR/issue labels reflect ready-to-ship-beta state after review.
|
||||
- [ ] Merge is skipped.
|
||||
|
||||
## Risk Assessment
|
||||
|
||||
Risk: GitHub auth lacks label/PR permissions.
|
||||
Mitigation: stop with exact `gh` error if label or PR creation fails.
|
||||
|
||||
## Security Considerations
|
||||
|
||||
Do not include secrets or private env output in GitHub comments/PR body.
|
||||
@@ -0,0 +1,50 @@
|
||||
---
|
||||
title: Add qwen3.7-plus to Bailian Coding provider
|
||||
description: ''
|
||||
status: pending
|
||||
priority: P2
|
||||
issue: 169
|
||||
branch: codex/issue-169-qwen37-bailian-model
|
||||
tags: []
|
||||
blockedBy: []
|
||||
blocks: []
|
||||
created: '2026-06-12T10:09:09.353Z'
|
||||
createdBy: 'ck:plan'
|
||||
source: skill
|
||||
---
|
||||
|
||||
# Add qwen3.7-plus to Bailian Coding provider
|
||||
|
||||
## Overview
|
||||
|
||||
Add `qwen3.7-plus` to GoClaw's hardcoded Bailian Coding model catalog so the shared `/v1/providers/{id}/models` endpoint exposes it to both web and desktop model pickers. Keep runtime request behavior backward-compatible: Bailian remains an OpenAI-compatible provider, and no GPT/Codex `reasoning_effort` metadata is added for this Qwen model.
|
||||
|
||||
## Phases
|
||||
|
||||
| Phase | Name | Status |
|
||||
|-------|------|--------|
|
||||
| 1 | [Catalog and capability docs](./phase-01-catalog-and-capability-docs.md) | Completed |
|
||||
| 2 | [Tests and verification](./phase-02-tests-and-verification.md) | Completed |
|
||||
| 3 | [Ship beta PR](./phase-03-ship-beta-pr.md) | Pending |
|
||||
|
||||
## Dependencies
|
||||
|
||||
- Source issue: <https://github.com/digitopvn/goclaw/issues/169>
|
||||
- Verified code paths:
|
||||
- `internal/http/provider_models_catalog.go`: `bailianModels()` hardcoded catalog.
|
||||
- `internal/http/provider_models.go`: Bailian dispatch uses `bailianModels()` then `withReasoningCapabilities()`.
|
||||
- `internal/http/providers.go`: Bailian runtime registration uses `providers.NewOpenAIProvider(...)`.
|
||||
- `internal/providers/openai_request.go`: `reasoning_effort` is gated by `LookupReasoningCapability()`, so do not add `qwen3.7-plus` there.
|
||||
|
||||
## Acceptance Criteria
|
||||
|
||||
- [x] `qwen3.7-plus` appears in the Bailian Coding model catalog.
|
||||
- [x] Display name is `Qwen 3.7 Plus`.
|
||||
- [x] Capabilities are documented as Text Generation, Deep Thinking, Visual Understanding.
|
||||
- [x] Web/Desktop provider model pickers inherit the model through the existing `/v1/providers/{id}/models` API.
|
||||
- [x] Existing Bailian provider registration remains backward-compatible.
|
||||
- [x] Tests cover the Bailian model list.
|
||||
|
||||
## Unresolved Questions
|
||||
|
||||
- Does Bailian expose explicit request parameters for Deep Thinking on `qwen3.7-plus`? Out of scope for this catalog update unless verified separately.
|
||||
@@ -0,0 +1,27 @@
|
||||
## Outcome
|
||||
Add `qwen3.7-plus` to the Bailian Coding provider model catalog so the shared provider model API exposes it to web and desktop model pickers.
|
||||
|
||||
## Implementation
|
||||
- Branch: `codex/issue-169-qwen37-bailian-model`
|
||||
- Plan: `plans/260612-1709-qwen37-bailian-model/plan.md`
|
||||
- Mode: `beta`
|
||||
- PR: pending
|
||||
|
||||
## Acceptance Criteria
|
||||
- [ ] `qwen3.7-plus` appears in Bailian Coding model catalog
|
||||
- [ ] Display name is `Qwen 3.7 Plus`
|
||||
- [ ] Capabilities are documented/represented as Text Generation, Deep Thinking, Visual Understanding
|
||||
- [ ] Web provider model picker can select `qwen3.7-plus` through `/v1/providers/{id}/models`
|
||||
- [ ] Desktop provider model picker can select `qwen3.7-plus` through the same backend model API
|
||||
- [ ] Existing Bailian provider registration remains backward-compatible
|
||||
- [ ] Tests cover the Bailian model catalog
|
||||
- [ ] Docs updated where provider model support is documented
|
||||
|
||||
## Pipeline State
|
||||
- [x] Worktree and branch created
|
||||
- [x] TDD plan created
|
||||
- [x] Plan validated
|
||||
- [x] Plan red-teamed
|
||||
- [ ] Cook complete
|
||||
- [ ] PR reviewed and fixed
|
||||
- [ ] Merged and CI green (only when --ship)
|
||||
+69
@@ -0,0 +1,69 @@
|
||||
## Code Review Summary
|
||||
|
||||
### Scope
|
||||
|
||||
- Files: `internal/http/provider_models_catalog.go`, `internal/http/provider_models_test.go`, `docs/02-providers.md`, `docs/12-extended-thinking.md`, `docs/project-changelog.md`
|
||||
- LOC: +76 / -0 pending diff
|
||||
- Focus: pending changes only, issue #169 Bailian Coding `qwen3.7-plus` catalog
|
||||
- Scout findings: `/v1/providers/{id}/models` dispatches Bailian to `bailianModels()`; web and desktop picker paths consume this endpoint; Bailian runtime registration remains `providers.NewOpenAIProvider(...)`; OpenAI `reasoning_effort` remains gated by GPT/Codex reasoning metadata and is not added for `qwen3.7-plus`.
|
||||
|
||||
### Overall Assessment
|
||||
|
||||
Implementation matches acceptance criteria. No blocking production-readiness issues found.
|
||||
|
||||
### Critical Issues
|
||||
|
||||
None.
|
||||
|
||||
### High Priority
|
||||
|
||||
None.
|
||||
|
||||
### Medium Priority
|
||||
|
||||
None.
|
||||
|
||||
### Low Priority
|
||||
|
||||
None.
|
||||
|
||||
### Edge Cases Found by Scout
|
||||
|
||||
- Verified Bailian catalog path requires API key before returning hardcoded models; test covers the current contract with a dummy key.
|
||||
- Verified UI propagation does not require hardcoded web/desktop model list updates because pickers call `/v1/providers/{id}/models`.
|
||||
- Verified `qwen3.7-plus` is not in `LookupReasoningCapability()`, so selecting it will not trigger unsupported OpenAI `reasoning_effort`.
|
||||
|
||||
### Positive Observations
|
||||
|
||||
- Test covers endpoint-level behavior, display name, and nil reasoning metadata.
|
||||
- Docs correctly distinguish advertised Bailian Deep Thinking capability from DashScope `enable_thinking` / `thinking_budget` request injection.
|
||||
- Existing Bailian provider default model and registration path are unchanged.
|
||||
|
||||
### Recommended Actions
|
||||
|
||||
1. No code changes required.
|
||||
2. Plan can mark Phase 2 complete; Phase 3 remains pending.
|
||||
|
||||
### Verification
|
||||
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/http -run 'TestProvidersHandlerListProviderModels(BailianIncludesQwen37Plus|ChatGPTOAuthIncludesReasoningMetadata|OpenAICompatAnnotatesKnownModels)'` passed.
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/providers -run 'TestLookupReasoningCapability|TestDashScopeModelSupportsThinking|TestDashScopeThinking'` passed.
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/http` passed.
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/providers` passed.
|
||||
- `git diff --check -- internal/http/provider_models_catalog.go internal/http/provider_models_test.go docs/02-providers.md docs/12-extended-thinking.md docs/project-changelog.md` passed.
|
||||
|
||||
### Metrics
|
||||
|
||||
- Type Coverage: N/A for Go; package compile covered by `go test`
|
||||
- Test Coverage: coverage percentage not collected
|
||||
- Linting Issues: 0 from `git diff --check`; full lint/vet not run
|
||||
|
||||
### Plan Status
|
||||
|
||||
- Phase 1: complete by diff evidence.
|
||||
- Phase 2: appears complete; focused and package tests passed.
|
||||
- Phase 3: pending, not reviewed as implementation.
|
||||
|
||||
### Unresolved Questions
|
||||
|
||||
None for this catalog implementation. Bailian-specific Deep Thinking wire controls remain explicitly out of scope.
|
||||
@@ -0,0 +1,19 @@
|
||||
# Scout Report
|
||||
|
||||
## Findings
|
||||
|
||||
- Repo verified: `digitopvn/goclaw`, default branch `dev`.
|
||||
- Source issue verified: #169 requests `qwen3.7-plus` for Bailian Coding.
|
||||
- Catalog source: `internal/http/provider_models_catalog.go` -> `bailianModels()`.
|
||||
- Endpoint dispatch: `internal/http/provider_models.go` uses `bailianModels()` for `provider_type == "bailian"`.
|
||||
- UI propagation: web and desktop model pickers call `/v1/providers/{id}/models`; no hardcoded Bailian model list found in UI.
|
||||
- Runtime boundary: `internal/http/providers.go` registers Bailian through `providers.NewOpenAIProvider(...)`.
|
||||
- Wire guardrail: `internal/providers/openai_request.go` sends OpenAI `reasoning_effort` only when `LookupReasoningCapability()` returns metadata or model is GPT/o-series.
|
||||
|
||||
## Decision
|
||||
|
||||
Implement catalog entry, endpoint regression test, and docs. Do not add `qwen3.7-plus` to the GPT/Codex reasoning capability registry in this task.
|
||||
|
||||
## Unresolved Questions
|
||||
|
||||
- Does Bailian expose explicit thinking request parameters for `qwen3.7-plus`? Not needed for the catalog issue.
|
||||
+19
@@ -0,0 +1,19 @@
|
||||
# Tester Report - issue #169 Bailian qwen3.7-plus catalog
|
||||
|
||||
## Scope
|
||||
- Verified the Bailian Coding model catalog addition in `internal/http/provider_models_catalog.go`.
|
||||
- Verified handler coverage in `internal/http/provider_models_test.go`.
|
||||
- Checked docs updates in `docs/02-providers.md`, `docs/12-extended-thinking.md`, and `docs/project-changelog.md`.
|
||||
|
||||
## Results
|
||||
- `PATH=/usr/local/go/bin:$PATH go test ./internal/http -run TestProvidersHandlerListProviderModelsBailianIncludesQwen37Plus -count=1`
|
||||
- PASS
|
||||
- `git diff --check`
|
||||
- PASS
|
||||
|
||||
## Notes
|
||||
- `GET /v1/providers/{id}/models` is the contract used by the web and desktop pickers; no UI hardcoded model list change was needed.
|
||||
- Bailian registration remains on the existing OpenAI-compatible provider path with the default model still set to `qwen3.5-plus`.
|
||||
|
||||
## Coverage gap
|
||||
- Catalog is covered by one focused regression only; no additional provider-level or UI-level test was added in this patch.
|
||||
@@ -0,0 +1,20 @@
|
||||
# Validation And Red-Team Report
|
||||
|
||||
## Validation
|
||||
|
||||
- Acceptance criteria are mapped to concrete files and tests.
|
||||
- Web/Desktop selection is covered through the shared backend model endpoint, not separate UI lists.
|
||||
- Backward compatibility is explicit: Bailian provider registration remains unchanged.
|
||||
|
||||
## Red-Team Findings
|
||||
|
||||
- Finding: Adding `qwen3.7-plus` to `LookupReasoningCapability()` would expose advanced reasoning controls and may make Bailian send OpenAI `reasoning_effort`.
|
||||
- Verdict: Reject for this issue scope. Keep runtime controls unchanged.
|
||||
- Finding: Vision capability is not a `ModelInfo` field today.
|
||||
- Verdict: Document capability only; avoid schema expansion for a single catalog entry.
|
||||
- Finding: Handler rejects providers with empty API keys before model dispatch.
|
||||
- Verdict: Test must create a Bailian provider with a dummy API key.
|
||||
|
||||
## Unresolved Questions
|
||||
|
||||
- None blocking implementation.
|
||||
Reference in new issue
Block a user