9 Commits

Author SHA1 Message Date
fe6666b665 fix: DBPools() not in __init__, sqlorContext uses module name not default 2026-05-08 11:41:53 +08:00
f9705f3892 fix: handle non-JSON LLM API responses gracefully
- Check resp.content_type before calling resp.json()
- Return clear error if server returns HTML (e.g. proxy error or wrong endpoint)
- Prevents 'Attempt to decode JSON with unexpected mimetype' crashes
2026-05-07 18:25:00 +08:00
e5e6796a4b feat: decrypt API key using ServerEnv.password_decode
- Add API key decryption in _resolve_provider using env.password_decode
- Matches project standard for encrypted credentials
2026-05-07 17:31:05 +08:00
1dbfd25772 fix: correct sor.R() call pattern - ns dict contains filters+sort
- sor.R signature: R(tablename, ns, filters=None)
- ns (2nd arg) is a combined dict: filter conditions + sort options
- No 3-arg calls or ns= keyword argument used
- All queries include user_id for multi-user isolation

Fixed in:
- llm_client.py: _get_llm_config uses sor.R(ns_dict)
- core.py: memory, sessions, skills, workflows, executions
- orchestrator.py: workflow tasks loading, execution recording
2026-05-07 17:08:29 +08:00
a88c964666 fix: multi-user isolation and sqlor-compatible queries
- Replace all sor.sqlExe(ORDER BY/LIMIT) with sor.R(ns={sort:...}) pattern
- Add user_id filter to all database queries for multi-user isolation
- Replace invalid sor.R(orderby=...) with ns={sort:...}
- _get_llm_config: use sor.R with user_id filter and ns sort
- core.py: fix _get_intelligent_memory_context, search_sessions,
  list_workflows, list_executions, manage_remote_skills (list)
- core.py: _update_memory_access_stats now requires user_id param
- orchestrator.py: fix _load_workflow_definition ns sort,
  _record_execution_end requires user_id param
- Add debug logging for config lookup (user_id, dbname, row data)
2026-05-07 16:39:07 +08:00
8e10a24d55 fix: LLM config query uses 'default' DB only, not calling module's DB
Problem: When integrated_crm_app calls llm_chat(), _get_llm_config() used
env.get_module_dbname('harnessed_agent') which returned 'crm_db' (the
calling module's DB), causing it to read stale/wrong config (qwen3-max)
instead of the correct config from default DB (qwen3.6-plus).

Fix: Remove get_module_dbname() call. LLM config is global and shared
across all modules, so always query from 'default' database.
2026-05-07 15:49:12 +08:00
6165c4b1cd fix: replace invalid sor.R(orderby/limit) calls with sqlExe
- All sor.R() calls with orderby= and/or limit= parameters replaced
  with raw SQL via sor.sqlExe() for correct execution
- Fixed in core.py (memory, sessions, workflows, executions, skills)
- Fixed in llm_client.py (config lookup)
2026-05-07 15:32:42 +08:00
a7f99fc710 fix: fallback to default DB for LLM config lookup
- _get_llm_config now tries module DB and 'default' DB
- Adds logging to show which DB is accessed
- Prevents empty config when table is in default DB
2026-05-07 15:10:53 +08:00
4e65ff8fe4 refactor: LLM client for calling supplier LLM APIs (not server)
Replaces wrong-direction llm_api.py (which served OpenAI endpoints)
with llm_client.py -- a client that calls supplier LLM APIs.

New module: llm_client.py
  - llm_chat(messages, model, temperature, ...) -> OpenAI response dict
  - llm_chat_stream(messages, ...) -> async generator of SSE chunks
  - llm_simple(prompt, system) -> plain text response
  - llm_list_models() -> list available models from provider
  - llm_get_config() -> show current config (key masked)
  - Supports provider presets: openai, dashscope, deepseek, siliconflow
  - Retry with exponential backoff (3 attempts)
  - 429 rate limit handling with Retry-After
  - Structured logging via appPublic.log

Model changes (harnessed_agent_config):
  - Add llm_provider (preset name: dashscope/openai/deepseek/siliconflow)
  - Add top_p field
  - llm_service_url defaults to DashScope compatible endpoint

Other:
  - Remove wrong-direction /v1/ endpoints
  - Fix pyproject.toml deps: only sqlor + bricks_for_python
  - Update init/data.json seed data with LLM config fields
  - Update CRUD view with llm_provider dropdown
2026-05-07 11:57:04 +08:00