LLM & ChatActive

RAG Chat GitHub

Chat with one branch of a GitHub repository: its files are indexed, and the closest matches go to the chat model you select. Each run needs a GitHub access token, public repositories included.

ChatLLMRAG

Model ID

rag-chat-github

Provider

wiro

Updated

Oct 24, 2026, 14:52

Support Wiro on G2

wiro playground—wiro/rag-chat-github

Select Base Model

Select a base model that will be run.

Select Private Base Model

Select a private base model that will be run.

Github Repo Owner Name

Enter a github respository owner name https://github.com/𝐦𝐞𝐭𝐚-𝐥𝐥𝐚𝐦𝐚/llama

Github Repo Name

Enter a github respository name https://github.com/meta-llama/𝐥𝐥𝐚𝐦𝐚

Github Branch Name

Enter a github respository branch name

Github Access Token

Provide your GitHub access token for private repositories. For public ones, leave it blank.

Prompt*

Prompt to send to the model.

Your request will cost $0.0006 per second.

Running this model on Wiro costs approximately $0.128072 in total.

(Total cost varies depending on the request's execution time.)

Sample outputs

Loading preview…

wiro-rag-chat-github-sample-1.txt

Updated Oct 24, 2026, 14:52

Overview

RAG Chat GitHub lets a chat model answer questions about a single branch of a GitHub repository. The app pulls the files on the branch you name into a search index of small chunks. The chunks closest to each question go to the model you selected.

What you can build

Onboarding assistants that walk new team members through an unfamiliar codebase
Q&A over a project's README and guides
Internal lookups that point to where a feature or config option is handled

Inputs

Repository owner and repository name. The defaults are meta-llama and llama.
Branch name, main by default.
A GitHub access token. Fill it in for every run, public repositories included.
The question. By default it asks for an explanation of what the repository does.
The answering model: an entry from Wiro's chat model list, or a private chat model on your account.
Optional fields for user and session IDs, which keep chat histories apart, plus a system prompt and sampling controls.

Outputs

A text answer. The sample reply for the default repository runs five sentences. It describes the repository as the home of the Llama 3.1 models and the Llama Stack, but that material comes from the deprecation notice at the top of its README, which points readers to newer repositories. meta-llama/llama itself is the deprecated Llama 2 repository, a minimal example for loading the models and running inference.

Recommended settings

Code splits poorly at small sizes. The 256-token default chunk can cut a function in half, so raise the chunk size when you ask about specific functions or classes, and keep some overlap (25 tokens by default) so text near a boundary lands in both neighbouring chunks. Five chunks are retrieved per question. More of them help with questions about how modules fit together, though they can also pull in unrelated files. For factual lookups, lower the temperature from 0.7 or turn sampling off.

Limitations

The whole branch is indexed. There is no option to skip folders or file types.
Only a branch can be named. There is no field for a tag or a specific commit.
Answers see only the retrieved chunks, so questions that need the full picture of a large codebase can come back incomplete.

API quick start

Run rag-chat-github with a single API call.

POST https://api.wiro.ai/v1/Run/wiro/rag-chat-github

{
  "prompt": "Can you explain what this repository does?",
  "user_id": "...",
  "session_id": "...",
  "selectedModel": "..."
}

curl

curl -X POST "https://api.wiro.ai/v1/Run/wiro/rag-chat-github" \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_WIRO_API_KEY" \
  --data-binary @- <<'JSON'
{
  "prompt": "Can you explain what this repository does?",
  "user_id": "...",
  "session_id": "...",
  "selectedModel": "..."
}
JSON