Notes re: configuration and use and generally what this is.

Readme.md

+# binary-bob
+
+Binary Bob knows everything about Synchronet.
+
+This is a RAG app that handles indexing of Synchronet support content and querying an LLM with relevant context.
+
+## Configuration
+
+The following environment variable *must* be set, and can optionally be in a `.env` file at the root of the project.
+
+```sh
+OPENAI_API_KEY=your-api-key-goes-here
+```
+
+The following optional environment variables are available, with defaults shown below:
+
+```sh
+OPENAI_EMBEDDING_MODEL=text-embedding-3-small
+EMBEDDING_DIMENSION=1536
+HTTP_PORT=3000
+LLM=gpt-3.5-turbo-0125
+QDRANT_COLLECTION=synchronet
+QDRANT_HOST=qdrant
+QDRANT_PORT=6333
+```
+
+## Use
+
+```sh
+docker compose up --build --force-recreate
+```
+
+## Notes
+
+I only intend to run this locally as a backend for the Chad Jipiti bot. Security of any kind has not been taken into account.
+
+Some work yet to be done on content ingestion, the current method is pretty sucky. I also need to add some more sources such as the Synchronet JS Object Model page and the legacy docs (Baja et al). Ultimately there will be some kind of task scheduler in the app or the host will kick off ingestion via cron and `docker exec`.
+
+`gpt-3.5-turbo-0125` isn't doing the best job, but it is at least very cheap. `gpt-4o` was producing pretty good responses, but is way too expensive for this project at the moment.

src/lib/config.ts

@@ -6,7 +6,7 @@ export default {
 	embeddingModel: process.env.OPENAI_EMBEDDING_MODEL ?? 'text-embedding-3-small',
 	embeddingDimension: parseInt(process.env.EMBEDDING_DIMENSION ?? '1536', 10),
 	httpPort: process.env.HTTP_PORT ?? 3000,
-	llm: 'gpt-3.5-turbo-0125',
-	qdrantCollection: 'synchronet',
+	llm: process.env.LLM ?? 'gpt-3.5-turbo-0125',
+	qdrantCollection: process.env.QDRANT_COLLECTION ?? 'synchronet',
 	qdrantUrl: `http://${process.env.QDRANT_HOST ?? 'qdrant'}:${process.env.QDRANT_PORT ?? '6333'}`,
 }
\ No newline at end of file