5dc739b800
feat: integration test suite — 416 tests, 61% coverage
...
Add OpenBao and Kratos to docker-compose dev stack with bootstrap
seeding. Full integration tests hitting real services:
- Vault SDK: KV read/write/delete, re-auth on bad token, new_with_token
constructor for dev mode
- Kratos SDK: list/get/create/disable/enable users, session listing
- Token store: PAT lifecycle with OpenBao backing, expiry handling
- Identity tools: full tool dispatch through Kratos admin API
- Gitea SDK: resolve_username, ensure_token (PAT auto-provisioning),
list/get repos, issues, comments, branches, file content
- Devtools: tool dispatch for all gitea_* tools against live Gitea
- Archive indexer: batch flush, periodic flush task, edit/redact/reaction
updates against OpenSearch
- Memory store: set/query/get_recent with user scoping in OpenSearch
- Room history: context retrieval by timestamp and event_id, access
control enforcement
- Search archive: keyword search with room/sender filters, room scoping
- Code search: language filter, repo filter, branch scoping
- Breadcrumbs: symbol retrieval, empty index handling, token budget
- Bridge: full event lifecycle mapping, request ID filtering
- Evaluator: DM/mention/silence short-circuits, LLM evaluation path,
reply-to-human suppression
- Agent registry: list/get_id, prompt reuse, prompt-change recreation
- Conversations: token tracking, multi-turn context recall, room
isolation
Bug fixes caught by tests:
- AgentRegistry in-memory cache skipped hash comparison on prompt change
- KratosClient::set_state sent bare PUT without traits (400 error)
- find_code_session returns None on NULL conversation_id
2026-03-24 14:34:03 +00:00
495c465a01
refactor: remove legacy responder + agent_ux, add Gitea integration tests
...
Legacy removal:
- DELETE src/brain/responder.rs (900 lines) — replaced by orchestrator
- DELETE src/agent_ux.rs (184 lines) — UX moved to transport bridges
- EXTRACT chat_blocking() to src/brain/chat.rs (standalone utility)
- sync.rs: uses ConversationRegistry directly (no responder)
- main.rs: holds ToolRegistry + Personality directly (no Responder wrapper)
- research.rs: progress updates via tracing (no AgentProgress)
Gitea integration testing:
- docker-compose: added Gitea service with healthcheck
- bootstrap-gitea.sh: creates admin, org, mirrors 6 real repos from
src.sunbeam.pt (sol, cli, proxy, storybook, admin-ui, mistralai-client-rs)
- PAT provisioning for SDK testing without Vault
- code_index/gitea.rs: fixed directory listing (direct API calls instead
of SDK's single-object parser), proper base64 file decoding
New integration tests:
- Gitea: list_repos, get_repo, get_file, directory listing, code indexing
- Web search: SearXNG query with result verification
- Conversation registry: lifecycle + send_message round-trip
- Evaluator: rule matching (DM, own message)
- gRPC bridge: event filtering, tool call mapping, thinking→status
2026-03-24 11:45:43 +00:00
b8b76687a5
feat(grpc): dev mode, agent prefix, system prompt, error UX
...
- gRPC dev_mode config: disables JWT auth, uses fixed dev identity
- Agent prefix (agents.agent_prefix): dev agents use "dev-sol-orchestrator"
to avoid colliding with production on shared Mistral accounts
- Coding sessions use instructions (system prompt + coding addendum)
with mistral-medium-latest for personality adherence
- Conversations API: don't send both model + agent_id (422 fix)
- GrpcState carries system_prompt + orchestrator_agent_id
- Session.end() keeps session active for reuse (not "ended")
- User messages posted as m.notice, assistant as m.text (role detection)
- History loaded from Matrix room on session resume
- Docker Compose local dev stack: OpenSearch 3 + Tuwunel + SearXNG
- Dev config: localhost URLs, dev_mode, opensearch-init.sh for ML setup
2026-03-23 17:07:50 +00:00
