Memory System
flyto-ai includes a persistent conversation memory system with full-text and semantic search.
Architecture
SQLite Store
├── Session management (conversations)
├── Message storage (full history)
├── BM25 index (keyword search)
├── Vector embeddings (semantic search)
└── Auto-summarization (context compaction)Features
Persistent Sessions
Conversations are stored in SQLite and survive restarts:
python
AgentConfig(
enable_memory=True,
memory_db_path="~/.flyto/memory.db",
)Search
Two search modes work together:
| Mode | How | Best For |
|---|---|---|
| BM25 | Keyword matching | Exact terms, function names |
| Embeddings | Semantic similarity | Conceptual queries |
Context Compaction
When the context window fills up, old messages are automatically summarized:
- Detect context approaching limit
- Summarize oldest messages into a compact summary
- Replace original messages with summary
- Continue conversation with full context
Auto-summarization
Long conversations are periodically summarized to maintain searchability without storing excessive tokens.
Configuration
| Option | Default | Description |
|---|---|---|
enable_memory | True | Enable persistent memory |
memory_db_path | ~/.flyto/memory.db | SQLite database path |
embedding_model | text-embedding-3-small | OpenAI embedding model |