feat: support Ollama with synthetic-data-kit

[rachittshah]

Ollama Backend Integration

Resolves #3

Changes

• Added dual backend system with abstract BaseLLMBackend class
• Implemented OllamaBackend with OpenAI-compatible API support
• Updated LLMClient to support backend selection (vLLM/Ollama)
• Added Ollama configuration in YAML with model, API, and retry settings
• Enhanced system-check command to display available Ollama models

Testing

• Tested with Ollama models: llama3.2, qwen3, phi3
• Verified compatibility with existing QA, CoT, and summary generation
• Confirmed model listing and availability checks

Documentation

• Updated README with Ollama setup and usage instructions
• Added backend-specific troubleshooting guides
• Added example commands for both vLLM and Ollama workflows

Migration

No breaking changes. Existing vLLM users can continue using the default backend.

[facebook-github-bot]

Hi @rachittshah!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[init27]

Thank you SO MUCH for the instant PR, Rachit! I will test and review later today/tomorrow

[rachittshah]

Sounds good Sanyam!

Let me know if there's any changes needed, and we can pickup SGLang next :D

[rachittshah]

@init27 did you get a chance to review this PR?

[dhuebner]

@init27
vLLM installation might become very challenging on some platforms. It would be really nice to have Ollama support soon. Thanks!

[init27]

Thanks for the follow up @rachittshah and @dhuebner!

I was planning to add CI/CD tests to main but it's taking a bit, so let me review the PR now please. Thanks again for the reminder!

[init27]

This is looking great, I did a quick review and test:

• We need to fix cli.py to work with ollama/vLLM to make curate functions work
• After this we need to re-test with curate function

[arty-hlr]

Out of curiosity, why focus on individual backends when it could support openAI compatible APIs instead, which would unlock a lot more backends, including ollama?

[init27]

@arty-hlr great question and valid point, I think the main reason being we want to focus on llama model supports.

IIRC, some of the model endpoints/providers are not OAI compatible and ideally we'd like to maximize support for them hence this approach. But open to your suggestions like always.

[dhuebner]

@init27 I just tried synthetic-data-kit with Ollama over api-endpoint and it worked well. Except of some problems when running in a Notebook, I was able to create and curate some data. Is the new version with api-endpoint already published?

[init27]

@dhuebner thanks for the feedback and trying! Can you please open an issue or share the details of the problems with notebook please, I can take a look.

I've put out an alpha version: https://pypi.org/project/synthetic-data-kit/0.0.4a0/

[rachittshah]

@init27 what do you need from me to merge this PR?