config/README.md

Project Configuration

.env File

This file contains the necessary API keys required for the application to function properly. Obtain the API keys from the following sources:

• OPENAI_API_KEY
• PINECONE_API_KEY
• AWS_ACCESS_KEY
• AWS_SECRET_ACCESS_KEY

config.json

This JSON file holds crucial configuration values for the entire application. Please refer to the documentation before modifying any configurations.

Pinecone Configuration

• PINECONE_INDEX_NAME: The name of the index, the highest-level organizational unit of vector data in Pinecone.
• PINECONE_VECTOR_DIMENSION: Dimensionality of the embedding model's vectors.
• PINECONE_UPSERT_BATCH_LIMIT: Number of transcript rows inserted into Pinecone Serverless in parallel.
• PINECONE_TOP_K_RESULTS: Number of results fetched by Pinecone for a query.
• PINECONE_DELTA_WINDOW: Conversation window size fetched for TOP_K results.
• PINECONE_CLOUD_PROVIDER: Cloud provider for Pinecone DB.
• PINECONE_REGION: Region of the Pinecone Cloud provider.
• PINECONE_METRIC: Distance metric used by Pinecone to calculate similarity.
• PINECONE_NAMESPACE: Logical separation inside the Pinecone Index.

Embedding Provider Configuration

• EMBEDDING_PROVIDER: Provider of the embedding model for text-to-vector conversion.
• EMBEDDING_MODEL_NAME: Name of the embedding model provided by the provider.

AWS Configuration

• AWS_INPUT_BUCKET: Bucket for storing audio files for AWS Transcribe.
• AWS_OUTPUT_BUCKET: Bucket collecting transcribed files.
• AWS_REGION: AWS region in use.
• AWS_TRANSCRIBE_JOB_NAME: Default name for Transcribe job.

LangChain Configuration

• LC_LLM_TEMPERATURE: Temperature value for the Large Language Model.
• LC_CONV_BUFFER_MEMORY_WINDOW: Conversation memory window limit. (Future Use)
• LC_LLM_SUMMARY_MAX_TOKEN_LIMIT: Maximum tokens allowed for summary in the memory buffer.
• LC_LLM_MODEL: Large Language Model used for inference.