Release 0.0.3

[eyurtsev]

In preparation for release 0.0.3 (likely mid or late April):

UI

• [ ] Add examples configuration to the UI
• [ ] Make sure examples work with the various model providers
• [ ] Add ability to edit system message

BE:

• [ ] Support download from a URL
• [ ] Expose patch/put endpoint so the UI can update the system message on an existing extractor
• [ ] Move to tool calling (for invocation and formatting of few-shot examples) and support Anthropic, Gemini, and Mistral (via langchain-mistralai).
• [ ] Expand support to content like python files which at the moment are probably not plain/text mimetype, but can be definitely parsed as is
• [ ] https://github.com/langchain-ai/langchain-extract/issues/122

[eyurtsev]

Review any other items from here: https://github.com/langchain-ai/langchain-extract/issues/25

[eyurtsev]

cc @ccurme

[eyurtsev]

Community members feel free to request / suggest features

[lhotwll217]

@eyurtsev what is download from URL? I've been messing around with an extract-from-url endpoint. Works decent considering I'm a noob

[eyurtsev]

Given a url, download the content, parse it and run it through extraction.

I want to augment the existing endpoint to accept a url in addition to text or a file.

The backend needs to: 1) have a way to enable/disable the ability to accept a url -- since folks need to configure it to use a proxy for it to be safe against ssrf attacks 2) information on whether it's enabled or disabled should be sent to the ui via the configuration endpoint

[danb27]

Request: download the results from the UI in JSON/CSV

[geraud-recarta]

Request: export/save an extractor Steps: create an extractor, improve via examples, export/save to use somewhere else Obviously import an extractor would be needed as well

[shima-khoshraftar]

Request: adding support for local models using ollama (ollamaFunctions) or any other library