llamabot
llamabot copied to clipboard
ericmjl
Pythonic class-based interface to LLMs
After installing llamabot in a fresh conda env with python 3.9 i get the error in the title. Any fixes recommended?
Incorporating BM25 is a good practice. BM25 is a good first retriever, from which we can use a vector database to further rank document chunks to include in context.
Use this issue to add in contributors. Example: ``` @all-contributors please add @andrewgiessel for ideas and code ```
I hear it's good! Should give it a shot, see how it works on a benchmark set of queries.
Some responses from `QueryBot` sound like it hasn't retrieved all the relevant chunks for answering the prompt. According to debugging, it retrieved only 4 `RetrievedMessage` instances instead of the configured...
[The `QueryBot` tutorial](https://ericmjl.github.io/llamabot/tutorials/querybot/) appears to be outdated. * It doesn't mention the `collection_name`. * Instead, it mentions an `index.json`. What are the steps to update it to the new API?...
* Sprint goals: * Getting started * Set up with Ollama * Set up development environment (devcontainer/codespace/conda env) * Beginner * Improve/help write documentation on: * Setting up development environment....
Provides an Ollama-compatible REST API to QueryBot. Example usage: ``` llamabot serve querybot --system-prompt "be smart" --collection-name \ junk1 --document-paths ~/h/duckdb-web/docs/sql/query_syntax/where.md ```
Dropping notes here. https://github.com/marhar/llmuri h/t @marhar
Originally observed this error in a [CI/CD pipeline run](https://github.com/ericmjl/llamabot/actions/runs/9284856152/job/25548132098#step:6:351). For this reason, I am pinning NumPy to