|
|
0d0c8299e8
|
fix: LIMIT $ → direct interpolation to avoid MySQL variable error
|
2026-08-02 15:46:37 +08:00 |
|
|
|
bb09bd36a0
|
refactor: use pipeline_conversations + pipeline_agent_settings (shared tables)
|
2026-08-02 00:00:21 +08:00 |
|
|
|
97e2c6b2f4
|
feat: Phase 1 — intent recognition, guided conversation, context bar
|
2026-08-01 23:36:29 +08:00 |
|
|
|
84df91d9eb
|
fix: model_id form reading fallback to select element
|
2026-08-01 21:57:04 +08:00 |
|
|
|
a45b4503eb
|
debug: add _call_llm logging for api_key/base/model tracing
|
2026-08-01 17:32:23 +08:00 |
|
|
|
ac2ff432ed
|
fix: remove RC4 decrypt from cockpit_chat, use plaintext api_key
|
2026-08-01 16:57:16 +08:00 |
|
|
|
328798f5a7
|
fix: tolerate plaintext api_keys on decrypt failure
|
2026-08-01 16:42:32 +08:00 |
|
|
|
13c466069f
|
fix: decrypt api_key before calling LLM in cockpit_chat
|
2026-08-01 16:32:46 +08:00 |
|
|
|
72cd262cdb
|
feat(cockpit): product-grade LLM integration with context + model mgmt
Backend:
- cockpit_chat.dspy: real LLM calls via llm table config, Hermes-style
context (system prompt + project/iteration + history), auto model
selection (multimodal if files, text otherwise)
- cockpit_model_options.dspy: returns active models with capabilities
- llm table: added capabilities column (text/vision/multimodal)
Frontend:
- inputed script: no page reload, updates agent bubble with LLM reply
- stats_row refreshed on success, error shown in bubble
DB:
- sd_agent_settings table for per-user agent preferences
|
2026-07-31 18:06:27 +08:00 |
|
|
|
a4ac438c49
|
fix: sor.I→sor.C for insert (I is inspect, C is create)
|
2026-07-31 15:01:03 +08:00 |
|
|
|
bd0890a394
|
feat(sdlc): cockpit driving mode — conversation timeline + agent chat
|
2026-07-30 15:53:18 +08:00 |
|
