Settings
models
chat_model
Chat/reasoning model name
embedding_model
Embedding model name
embedding_dimensions
Embedding vector dimensions
chat_max_tokens
Max tokens for chat response
chat_temperature
Temperature for chat model
vector_db
vector_store
Active vector store: pgvector | pinecone | qdrant
pgvector_logs_collection
PGVector collection for container logs
vector_ttl_days
Days to keep log vectors before auto-cleanup (0=disabled)
chunking
chunk_size
Text chunk size in characters
chunk_overlap
Overlap between chunks
chunk_strategy
Chunking strategy: auto | recursive | semantic | log
semantic_breakpoint_type
Semantic chunker breakpoint type: percentile | standard_deviation | interquartile
semantic_breakpoint_threshold
Semantic chunker breakpoint threshold (percentile value or multiplier)
log_chunk_size
Chunk size for log documents (characters)
log_chunk_overlap
Overlap for log document chunks
chat
chat_system_prompt
System prompt for the AI assistant
training
auto_training_enabled
Enable nightly auto-training
auto_training_cron
Cron schedule for auto-training (UTC)
rerank
rerank_enabled
Enable reranking of retrieved documents
rerank_model
Reranking model name
rerank_top_n
Number of documents to keep after reranking
alerting
alert_error_threshold
Min errors in a log batch to trigger an alert
alert_webhook_url
Webhook URL for error alerts (Teams/Slack)
Vector Store Connection
These are set via environment variables. To change, update the deployment config and restart.
PGVector Host
Set via PGVECTOR_HOST env var
Pinecone API Key
Set via PINECONE_API_KEY env var
Qdrant URL
Set via QDRANT_URL env var
LLM API Key
Set via LLM_API_KEY env var
Danger Zone
Delete All Chunks
Permanently remove all vector embeddings from all collections and clear document records.