Prompt file imported from Bryan-Roe/Aria (
.github/prompts/chat.prompt.md). Fill in{{input}}before use. Copyright stays with the author.
Chat
Start a chat conversation using the Aria platform's multi-provider system with semantic memory.
Pipeline
- Provider auto-detection (
detect_provider()): LM Studio → Ollama → Azure OpenAI → OpenAI → Local (LoRA explicit-only) - Memory injection: Embed user message → fetch similar past messages → inject as context
- Token pruning: Fit conversation within context window budget
- Streaming: SSE-based response delivery with movement tag extraction
- Self-learning: Log conversation to JSONL for future training
Provider Configuration
| Provider | Required Env Vars |
|---|---|
| Azure OpenAI | AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_DEPLOYMENT, AZURE_OPENAI_API_VERSION |
| OpenAI | OPENAI_API_KEY |
| LMStudio | LMSTUDIO_BASE_URL |
| LoRA | --provider lora + adapter path |
| Local | None (zero-dependency fallback) |
Memory Modes
- Semantic memory: Cosine similarity search on embeddings
- Session memory: Scoped to current session_id
- No memory: Stateless single-turn
Usage
Chat with: {{input}}
Using provider: [auto-detect | azure | openai | lmstudio | lora | local]
Memory: [enabled | disabled]
Streaming: [yes | no]
Endpoints
POST /api/chat— Standard chat with memory injectionPOST /api/chat/stream— SSE streaming with movement commandsGET /api/ai/status— Check active provider and system health