{"repo":"Ben-Epstein/extract-anything","free":true,"listed":false,"github":"https://github.com/Ben-Epstein/extract-anything","clone":"git clone https://github.com/Ben-Epstein/extract-anything.git","description":"Adaptation of baml's extract-anything repo using modal for self-hosting","language":"TypeScript","stars":10,"topics":[],"license":"Apache-2.0","category":"self-hosted-apps","readme_excerpt":"Extract Anything Adapted from https://github.com/BoundaryML/baml-examples/tree/hellovai/extract-anything A ready-to-go repo for building complex extraction systems: BAML for bulletproof LLM extractions with type-safe outputs Prefect for orchestrating extraction jobs from raw PDF - extraction - Delta table for analytics Modal for compute of prefect jobs + deploying open source LLMs for processing Delta lake for storing extracted results for downstream analytics and data processing GitHub CI/CD Full CI/CD workflows Infrastructure Dataflow This repo comes equipt to run prefect jobs on modal, processing raw documents with baml. To configure prefect and modal, follow this guide. See the Makefile for some helpful aliases. Required repo secrets MODAL TOKEN ID : https://modal.com/settings/[YOUR ACCOUNT]/member-tokens MODAL TOKEN SECRET : https://modal.com/settings/[YOUR ACCOUNT]/member-tokens SLACK BOT TOKEN : https://api.slack.com/tutorials/tracks/getting-a-token PREFECT API KEY : See the output of make show-prefect-profile (after running make prefect-auth ) PREFECT API URL : See the output of make show-prefect-profile (after running make prefect-auth ) Requred .env values This depends on what LLM you're using. If OpenAI, then the standard OPENAI API KEY, if vertex, then the path to your vertex-json credentials. If you are using LLMs deployed in modal, you need to set LLM BASE URL See the .env.example for examples.","default_branch":null,"files":null,"tree":[],"storefront":"/r/Ben-Epstein","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/Ben-Epstein/extract-anything/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}