Vex — Vector Exchange v0.6.1

# Vex — Vector Exchange v0.6.1

> Cross-standard vector DB migration tool with cryptographic memory portability. Export, sign, verify, and migrate agent memory between vector stores using the open `.vmig.jsonl` interchange format.

https://medium.com/ai-in-plain-english/your-agent-memory-is-trapped-heres-the-key-87224b483534
<img width="720" height="514" alt="image" src="https://github.com/user-attachments/assets/cac0e561-7fae-4853-9c9f-9960a78d6f2f" />



<img width="1086" height="470" alt="image" src="https://github.com/user-attachments/assets/ed5cccb1-40b4-4164-a190-ec4478150a82" />





```bash
npx @vektormemory/vex export --from vektor --db slipstream-memory.db --output memories.vmig.jsonl
npx @vektormemory/vex sign memories.vmig.jsonl
npx @vektormemory/vex verify memories.vmig.jsonl
npx @vektormemory/vex import --from memories.vmig.jsonl --to pinecone --api-key $KEY --index my-index --host $HOST
npx @vektormemory/vex migrate --from vektor --to qdrant --db memory.db --url http://localhost:6333 --collection memories
```

# Why

Every vector DB has a different API, a different format, and zero interop. Every LLM platform locks your conversation history behind a proprietary export. Moving your agent memory between systems means writing a one-off script every time.

`vex` fixes that with a single open format, a growing connector library, and a set of format adapters that turn your conversation history into whatever shape any LLM provider needs. Your memory is always exportable, always portable, always verifiable, always yours.

---

# What's New in v0.6.1

**Conversation portability.** Import your entire Claude or ChatGPT history into any vector DB in one command. Convert it to OpenAI fine-tuning format, Anthropic Messages API, Groq, Perplexity, Mistral, or plain text. Your conversations are now first-class agent memory.

- **`claude-export` connector** — import from the official claude.ai data export. Handles both claude.ai and API shapes. Three chunking modes: `turn`, `conversation`, `exchange`.
- **`chatgpt-export` connector** — import from the official ChatGPT data export. Reconstructs the active conversation thread from ChatGPT's tree-structured mapping format.
- **`vex convert` command** — transform `.vmig.jsonl` into provider-specific formats for fine-tuning or context injection. Five adapters: `openai-finetune`, `openai-context`, `generic-chat`, `anthropic-finetune`, `plain-text`.
- **Schema-adaptive `vektor` connector** — `load()` now detects the target DB's column names, primary key type, and schema version at runtime. Works with any VEKTOR DB — minimal, full SDK, integer ID, TEXT ID — without modification.
- **Accurate import counters** — summary numbers now come from the connector's own upserted/skipped counts, not a vector-presence heuristic. Text-only imports (no vectors) show the correct count.

---

# Connectors

| Connector | Export | Import | Notes |
|-----------|--------|--------|-------|
| `vektor` | ✅ | ✅ | VEKTOR Slipstream SQLite — schema-adaptive, sqlite-vec ANN optional |
| `jsonl` | ✅ | ✅ | `.vmig.jsonl` round-trip |
| `pinecone` | ✅ | ✅ | Tested 4,900 vectors |
| `qdrant` | ✅ | ✅ | Tested 3,917 vectors, auto-create collection |
| `chroma` | ✅ | ✅ | Auto-create collection |
| `weaviate` | ✅ | ✅ | GraphQL cursor pagination, extractStream |
| `pgvector` | ✅ | ✅ | Schema introspection, extractStream |
| `claude-export` | ✅ | — | Claude conversation JSON → vmig |
| `chatgpt-export` | ✅ | — | ChatGPT conversation JSON → vmig |

---

# Convert Adapters

| Adapter | Output | Use |
|---------|--------|-----|
| `openai-finetune` | `.jsonl` | Upload to `POST /v1/files` → `/v1/fine_tuning/jobs` |
| `openai-context` | `.json` | Inject as context into any chat completion call |
| `generic-chat` | `.jsonl` | Perplexity, Groq, Mistral, Together, Fireworks, Cerebras |
| `anthropic-finetune` | `.jsonl` | Anthropic Messages API format |
| `plain-text` | `.txt` | Human-readable transcript |

Aliases: `perplexity` / `groq` / `mistral` / `together` → `generic-chat` · `anthropic` → `anthropic-finetune` · `txt` → `plain-text`

---

# Install

```bash
npm install -g @vektormemory/vex

# Or run without installing
npx @vektormemory/vex --help

# Signing support (BLAKE3 + Ed25519)
npm install @noble/hashes @noble/ed25519

# Vex Adapter (vec2vec projection — optional, premium)
npm install @vektormemory/vex-adapter
```

Requirements: Node.js >= 18. No extra dependencies for Pinecone, Qdrant, Chroma, or Weaviate — connectors use the built-in fetch API. pgvector requires `npm install pg`.

---

## [0.6.1] - 2026-06-05

### Fixed
- vektor connector: FTS5 triggers with TEXT id as rowid caused datatype mismatch on import into Slipstream DBs with v1.5+ schema. Triggers suspended before bulk import and rebuilt after.
- vektor connector: created_at normalised to unix timestamp when target schema uses NUM column.

# Commands

## Conversation Export (v0.6.1)

```bash
# Claude — export from claude.ai Settings → Privacy 

... (truncated)