Commit Graph
4 Commits
Author SHA1 Message Date
Aska 3f9c4ca3c3 fix(prompt): prevent LLM from confusing context with user input
- Context merged into single system message with clear delimiters
- Prefix: 「以下是系统上下文,不是用户说的话」
- Memory → 【相关记忆】, IoT → 【设备状态】
- Filter empty memories, skip raw intent data
- Persona rule: context markers are internal data, not user input
2026-08-12 19:28:32 +08:00
Aska 585dfa1e38 fix(prompt): clean persona system prompt — remove YAML dump, filter empty subsessions
- Persona injector: concise natural language rules, not raw YAML
- Context builder: skip empty memory, skip raw intent data
- Prompt reduced from 2000+ tokens to ~500 tokens
- LLM response quality improved (natural tone, no confusion)
2026-08-12 19:27:12 +08:00
Aska 4e2e81f684 feat(trace): waterfall visualization + persona YAML sync
- trace.Logger: Trace+Span model with 14+ hop tracking
- trace.SpanTimer: convenience API for instrumentation
- trace_adapter.go: serialize Trace+Span for frontend
- Frontend TracePage: colored waterfall bars, expand/collapse, detail popup
- persona.go: full v1 YAML parser (identity/traits/forbidden/sections)
- cyrene_persona.yaml: v1 persona config copied to v2
- Orchestrator: TraceHook + MemoryExtractor injection points
2026-08-12 18:52:39 +08:00
Aska 3b3e5a0697 feat(ai-core): add LLM orchestration engine
- model: shared data types (Message, Intent, StreamToken, LLMMessage)
- persona: YAML/Raw loading + system prompt injection
- llm: OpenAI-compatible adapter with streaming + semantic cache
- context: sliding window context store with token budgeting
- subsession: parallel provider dispatch + review message parser
- orchestrator: intent analysis pipeline (IoT/greeting/chat/memory)
- background: autonomous thinking engine
- 20+ tests, build + vet clean, ~130 total tests
2026-08-09 21:21:27 +08:00