Auto Fine-Tuning

Describe your retrieval domain and prepare an embedding model adaptation workflow.

What is Auto Fine-Tuning?

Auto Fine-Tuning adapts an embedding model to a retrieval domain using descriptions and generated training examples. It is separate from training a classifier with labeled examples.

Embedding model adaptation

This configuration describes an Auto Fine-Tuning workflow. Chatsax does not currently expose a verified upstream submission API for it. You can prepare and export a configuration; no training job is started here.

Train a classifier with labeled examples

Query descriptionDocument description

Or, webpage URL

Webpage URL

Or, general instruction

General instruction

Jina — Base embedding model

BAAI — Base embedding model

Configuration preview

{
  "embedding_provider": "Jina",
  "base_model": "jina-embeddings-v5-omni-nano",
  "query_description": "",
  "document_description": ""
}

Embedding adaptation workflow — Source: Jina AI

But does it work though?

Published Jina experiments compare a base embedding model with its adapted model. These are reference results, not jobs run by Chatsax, and some tasks show a regression.

Dataset	Base model	Before	After	Change	Examples	Minutes	Artifacts
tollefj/norwegian-nli-triplets	jinaai/jina-embeddings-v2-base-en	0.8524	0.8671	1.7%	664	68	Dataset Model 1
mteb/askubuntudupquestions-reranking	jinaai/jina-embeddings-v2-base-en	0.6984	0.7221	3.4%	88	8	Dataset Model 2
mteb/scidocs-reranking	jinaai/jina-embeddings-v2-base-en	0.7726	0.8222	6.4%	88	9	Dataset Model 3
C-MTEB/CMedQAv2-reranking	jinaai/jina-embeddings-v2-base-zh	0.9381	0.9476	1.0%	88	9	Dataset Model 4
nc33/triplet_sbert_law2 (machine-translated to dutch)	jinaai/jina-embeddings-v2-base-en	0.9040	0.9483	4.9%	1304	210	Dataset Model 5
mteb/stackoverflowdupquestions-reranking	jinaai/jina-embeddings-v2-base-code	0.6397	0.6214	-2.9%	88	27	Dataset Model 6
mteb/stackoverflowdupquestions-reranking	jinaai/jina-embeddings-v2-base-code	0.6397	0.6193	-3.2%	88	11	Dataset Model 7
mteb/askubuntudupquestions-reranking	jinaai/jina-embeddings-v2-base-en	0.6449	0.6496	0.7%	88	12	Dataset Model 8

Source: Jina AI

Auto Fine-Tuning API

The configuration above is exportable, but a verified upstream Auto Fine-Tuning submission endpoint is not available in Chatsax. Explore the published models or use the separately supported classifier training API.

Published models Classifier training API

Rate limits

Your key’s assigned limits apply alongside the provider’s service capacity. Check the dashboard for the limits currently active on your account.

Service	Method	Usage basis
Reader	GET / POST	Output tokens
Search	GET / POST	Search request and content
Embeddings	POST	Input tokens and modalities
Reranker	POST	Query and candidate documents

View account limits Upstream rate reference

Frequently asked questions

API questions

Why is the first request for some models slow?

Some hosted models may need to load or initialize before serving a request. Latency also depends on the input size and current service load. Keep a timeout suited to the workload and inspect errors before retrying.

Can one Chatsax key access several APIs?

A key can access the services assigned to its scopes. Review the key in the dashboard before sharing it with an application.

Where can I review my usage?

Open the key manager to inspect recorded requests, status, latency, and available usage data for the selected key.

What if I lose a key?

Do keys expire?

Access follows the expiry and status configured for the key. A key that is revoked or expired cannot be used for new requests.

Can tokens be moved between keys?

Use only the transfer options available in your account. Do not assume that provider balances and Chatsax balances are interchangeable.

Can I revoke an API key?

Use the key manager. Revocation stops future access, so update any dependent applications before retiring a key.

Are my documents used to train models?

Check the platform privacy terms and the selected provider’s policy. Avoid sending material your organization has not approved for that service.

How do I handle rate limits?

Use the status and retry information returned by the API. Add backoff and control concurrency instead of continuously retrying the same request.

Can I process a large collection asynchronously?

The batch workflow supports submitting a job, checking its status, and downloading its output when the selected provider completes it.

Is Chatsax operated by Jina AI?

Chatsax is independently operated. The current model APIs are provided by Jina, whose papers and model cards are linked for reference.

How do I get an API key?

Open the dashboard to receive a visitor API key automatically. Copy it into your application or try a request immediately. Sign in when you want to manage an account and add tokens.

Get an API key

Billing questions

Is usage counted by sentences or requests?

Each service has its own metering rules. Reading, search, embedding, and generation can count different units; review the active pricing before use.

How do I get started?

Open the dashboard to receive a visitor key automatically. Sign in when you need account management or token purchases. The account screen shows your available balance and access.

Can an interrupted request consume resources?

The provider may have processed work before the connection ended. Review the recorded usage and billing status rather than inferring them from a closed browser tab.

Which payment methods are available?

The checkout screen displays payment methods available for your region, currency, and account.

Where can I find a receipt or invoice?

Use the billing portal linked from your account, and retain the payment confirmation associated with the purchase.

Does buying API credit grant a model license?

No. API credit pays for hosted usage. Downloading and commercially running model weights follows the model’s separate license.

Can I discuss a larger deployment?

Use the contact page to describe the services, traffic, latency, and deployment requirements you need.

Why is the balance different from what I expected?

Check the selected account and key, the payment status, and the latest usage records. Keep the payment reference when contacting support.

How do I manage future charges?

Use the billing controls provided in the account or payment portal. Check the confirmation after changing an automatic payment setting.