# Get started with Talqora Vector

Use Talqora when you need durable, filterable retrieval over vectors, sparse
text, or processed files. This guide is intentionally short so an agent can
fetch it before making a change.

## 1. Install an SDK

```bash
pip install talqora
# or: npm install talqora
```

## 2. Create an index

Create a Developer workspace and a narrowly scoped `tq_live_` API key in the
console. Keep the key on a server, never in browser code. The API defaults an
index to `us-east-1`, 1536 dimensions, and cosine distance when those options
are omitted.

```python
from talqora import Talqora

client = Talqora(api_key="tq_live_...")
index = client.indexes.create(name="support-search")
```

## 3. Send a file or write vectors

For a public or time-limited object URL, Talqora downloads, validates, and
processes the file asynchronously. Metadata is filterable retrieval context;
do not put a semantic question into metadata.

```python
job = client.files.import_from_url(
    index.id,
    source_url="https://storage.example.com/manual.pdf",
    metadata={"tenant": "acme", "source": "handbook"},
)
print(job.id, job.status)
```

## 4. Retrieve with least privilege

Use a read-only key scoped to the target index and include the authorization
boundary as a metadata filter. Dense and hybrid queries accept normal text;
Talqora creates the embedding when `query_text` is supplied.

```python
results = client.vectors.query(
    index.id,
    search_type="hybrid",
    query_text="what is the reimbursement policy?",
    filter={"tenant": "acme"},
    top_k=10,
    include_metadata=True,
)
```

## References

- OpenAPI: https://talqora.com/openapi.json
- Agent instructions: https://talqora.com/llms.txt
- Developer portal: https://talqora.com/developers
- MCP discovery: https://talqora.com/.well-known/mcp
- Full docs: https://docs.talqora.com