Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
GIT_TOKENNoHTTPS git personal access token
LLM_MODELNoChat modelgemma4:26b
NEO4J_URINoBolt URIbolt://localhost:7687
MCP_KB_ENVNoEnvironment namelocal
NEO4J_USERNoNeo4j usernameneo4j
QDRANT_URLNoQdrant REST endpointhttp://localhost:6333
GIT_USERNAMENoHTTPS git username
LLM_PROVIDERNoollama / openai / azureollama
OLLAMA_MODELNoOverride for enrichment model if LLM_MODEL is not used
OTEL_ENABLEDNoOpenTelemetry tool tracing/metricsfalse
POSTGRES_DSNNoPostgres URLlocal mcpkb
RBAC_ENABLEDNoRBAC enabledfalse
RERANK_MODELNoReranker idcross-encoder/ms-marco-MiniLM-L-6-v2
SPARSE_MODELNoFastEmbed sparse modelQdrant/bm25
EMBEDDING_DIMNoVector size (must match model)384
GRAPH_BACKENDNoRequired 'neo4j'neo4j
LANGFUSE_HOSTNoSelf-hosted or cloud Langfusehttp://localhost:3000
OTEL_EXPORTERNoconsole or otlpconsole
RALLY_API_KEYNoRally API key for story ids
NEO4J_DATABASENoDatabase nameneo4j
NEO4J_PASSWORDNoNeo4j passwordneo4j
OPENAI_API_KEYNoHosted LLM API key
QDRANT_API_KEYNoQdrant API key
RALLY_BASE_URLNoRally API base URLRally v2 URL
RERANK_ENABLEDNoLocal cross-encoder rerankfalse
SPARSE_ENABLEDNoQdrant BM25 sparse vectorstrue
TAVILY_API_KEYNoOptional Tavily; else ddgs
EMBEDDING_MODELNoEmbedding model idBAAI/bge-small-en-v1.5
OLLAMA_BASE_URLNoOpenAI-compatible Ollamahttp://localhost:11434/v1
OPENAI_BASE_URLNoHosted LLM base URL
LANGFUSE_ENABLEDNoLLM-level tracing (prompts/completions/latency)false
LEARNING_BACKENDNofile (JSON) or neo4j_qdrant (durable graph + payload boosts)file
LEARNING_ENABLEDNoOnline aliases + retrieval boosts from usagetrue
MCP_KB_HTTP_HOSTNoHTTP bind host127.0.0.1
MCP_KB_HTTP_PORTNoHTTP bind port8000
MCP_KB_LOG_LEVELNoLog levelINFO
MCP_KB_TRANSPORTNostdio or httpstdio
POSTGRES_ENABLEDNofalse = JSON metadata fallbacktrue
AUDIT_LOG_ENABLEDNoAudit log enabledtrue
MCP_KB_REPOS_ROOTNoCheckout / index root./data/repos
PROMPT_GUARD_MODENosanitize (neutralize) or drop (exclude the chunk)sanitize
EMBEDDING_PROVIDERNofastembed / sentence-transformers / openaifastembed
GIT_DEFAULT_BRANCHNoDefault branchmain
NEO4J_AUTH_ENABLEDNoSet false for passwordless local Dockertrue
RERANK_CANDIDATE_KNoRerank candidate pool30
WEB_SEARCH_ENABLEDNoQuery-time web gap-filltrue
LANGFUSE_PUBLIC_KEYNoLangfuse public key
LANGFUSE_SECRET_KEYNoLangfuse secret key
LEARNING_MODEL_PATHNoLightGBM model file location./data/models/ranker.txt
EMBEDDING_BATCH_SIZENoEmbed batch64
PROMPT_GUARD_ENABLEDNoScan retrieved context for prompt-injection phrasing before it reaches the LLMtrue
MCP_KB_REPOS_MANIFESTNoRepo list./config/repos.yaml
LEARNING_MODEL_MIN_AUCNoGuardrail — model must beat this held-out AUC to be saved0.55
LEARNING_TRAIN_ENABLEDNoEnable learning_cron.py scheduled retrainingfalse
OTEL_EXPORTER_ENDPOINTNoOTLP collector endpoint
SENSITIVE_SCAN_ENABLEDNoRedact secrets at index timetrue
LEARNING_DASHBOARD_PORTNoPort for learning dashboard8060
LEARNING_MODEL_BOOST_CAPNoMax +/- boost the model can contribute0.5
QDRANT_COLLECTION_PREFIXNoCollection prefixmcpkb
WEB_SEARCH_MIN_CONFIDENCENoWeb when pack confidence is below this0.45
LEARNING_MIN_TRAINING_ROWSNoMinimum aggregated rows before training is attempted30
LEARNING_DASHBOARD_AUTOSTARTNomcp-kb-server auto-starts cron trainer + dashboard as background threadsfalse
LEARNING_DASHBOARD_REFRESH_SNoDashboard auto-refresh interval15
LEARNING_TRAIN_INTERVAL_MINUTESNoRetraining cadence for the cron loop1

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Server capabilities have not been inspected yet.

Tools

Functions exposed to the LLM to take actions

NameDescription

No tools

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources