REST API and refactoring

Author: Kaltsoon

.env.template

@@ -5,4 +5,5 @@ POSTGRES_PASSWORD=postgres
 POSTGRES_DB=postgres
 REDIS_URL=redis://redis:6379
 OLLAMA_URL=http://host.docker.internal:11434
-OLLAMA_MODEL=llama3
+OLLAMA_MODEL=llama3
+API_PORT=3000

.github/workflows/test.yml

@@ -20,7 +20,7 @@ jobs:
           node-version: 20
 
       - name: Install dependencies
-        run: npm ci
+        run: npm ci --legacy-peer-deps
 
       - name: Compile TypeScript
         run: npx tsc --noEmit

Dockerfile

@@ -1,13 +1,15 @@
-FROM node:20
+FROM node:24
 
 WORKDIR /app
 
 COPY package*.json ./
 
-RUN npm install
+RUN npm install --legacy-peer-deps
 
 COPY . .
 
 RUN npm run build
 
-ENTRYPOINT ["npm", "run", "start"]
+EXPOSE 3000
+
+CMD ["npm", "run", "start:api"]

README.md

@@ -2,7 +2,7 @@
 
 [![Test](https://github.com/Kaltsoon/rag-my-docs/actions/workflows/test.yml/badge.svg)](https://github.com/Kaltsoon/rag-my-docs/actions/workflows/test.yml)
 
-A CLI application for building a [RAG](https://blogs.nvidia.com/blog/what-is-retrieval-augmented-generation/) for GitHub repository's markdown files and using it as a context for an LLM to ask questions related to the repository's documentation.
+A REST API and CLI application for building a [RAG](https://blogs.nvidia.com/blog/what-is-retrieval-augmented-generation/) for GitHub repository's markdown files and using it as a context for an LLM to ask questions related to the repository's documentation.
 
 The client code is implemented using TypeScript. PostgreSQL with [pgvector](https://github.com/pgvector/pgvector) is used as a vector database. Redis is used for caching purposes.
 
@@ -14,5 +14,29 @@ Docker and [Ollama](https://ollama.com/).
 
 1. Setup [Ollama](https://ollama.com/) and pull the `llama3:8b` model by running `ollama pull llama3:8b`. You can use other Ollama models as well, see next step for configuration instructions.
 2. Create a `.env` and with the contents of the `.env.template` file. Feel free to change the environment variables, e.g. `OLLAMA_MODEL` variable to change the model.
-2. Run the containers by running `docker compose up -d pg redis`.
-3. Once the containers are ready, run the CLI application by running `docker compose run -it --rm cli` and follow the instructions provided by the application.
+3. Run the containers by running `docker compose up -d`
+4. Once the containers are ready, there's two ways to access the LLM: `http://localhost:3000/api/repos/{owner}/{repo}/prompt` REST API endpoint and the CLI application. Details for both are below.
+
+> [!WARNING]  
+> GitHub's API [limits the requests for unauthenticated users](https://docs.github.com/en/rest/using-the-rest-api/rate-limits-for-the-rest-api?apiVersion=2022-11-28). Provide a GitHub authentication token as instructed below to increate your rate limit.
+
+### REST API
+
+REST API endpoint `http://localhost:3000/api/repos/{owner}/{repo}/prompt` (e.g. <http://localhost:3000/api/repos/Kaltsoon/rag-my-docs>) accepts a request body in the following format:
+
+```json
+{
+  "question": "What is the purpose of the repository?",
+  "ref": "dev",
+  "auth": "supersecret"
+}
+```
+
+The `ref` and `auth` attributes are optional. The `ref` attribute determines name of the commit/branch/tag. Defaults to repository's default branch. The `auth` attribute determines the GitHub authentication token used for requests to GitHub's API. It is only required while accessing non-public resources or to increase the rate limit.
+
+> [!WARNING]  
+> The first request for the repository will take some time, but the subsequent request to the same repository are much faster due to caching.
+
+### CLI
+
+You can run the CLI application by running `docker compose exec -it app npm run start:cli`. The CLI application will provide instructions on the usage.

compose.yml

@@ -1,9 +1,11 @@
 services:
-  cli:
+  app:
     build:
       context: ./
     env_file: .env
-    container_name: rag-my-docs-cli
+    container_name: rag-my-docs-app
+    ports:
+      - 3000:3000
   redis:
     image: redis/redis-stack-server
     volumes:
@@ -20,5 +22,4 @@ services:
     container_name: rag-my-docs-postgres
 volumes:
   redis_data:
-  ollama_data:
   pg_data: