AI Provider Configuration

Configure AI model providers including OpenAI, Anthropic, Google, Azure, and more

LibreChat supports multiple AI providers. Configure them using environment variables and the librechat.yaml file.

OpenAI

OPENAI_API_KEY · string · required

OpenAI API key

OPENAI_API_KEY=user_provided

Set to user_provided to allow users to provide their own keys

OPENAI_MODELS · string

Comma-separated list of available OpenAI models

OPENAI_MODELS=gpt-5,gpt-5-codex,gpt-5-mini,o3-pro,o3,o4-mini,gpt-4o,gpt-4o-mini

OPENAI_REVERSE_PROXY · string

Custom reverse proxy URL for OpenAI API

OPENAI_ORGANIZATION · string

OpenAI organization ID

DEBUG_OPENAI · boolean · default: "false"

Enable OpenAI debug logging

Conversation Titles

TITLE_CONVO · boolean · default: "true"

Enable automatic conversation title generation

OPENAI_TITLE_MODEL · string · default: "gpt-4o-mini"

Model used for generating conversation titles

Summarization

OPENAI_SUMMARIZE · boolean

Enable conversation summarization

OPENAI_SUMMARY_MODEL · string · default: "gpt-4o-mini"

Model used for conversation summarization

Anthropic (Claude)

ANTHROPIC_API_KEY · string · required

Anthropic API key

ANTHROPIC_API_KEY=user_provided

ANTHROPIC_MODELS · string

Comma-separated list of Anthropic models

ANTHROPIC_MODELS=claude-sonnet-4-6,claude-opus-4-6,claude-opus-4-20250514,claude-3-7-sonnet-20250219,claude-3-5-sonnet-20241022,claude-3-5-haiku-20241022

ANTHROPIC_REVERSE_PROXY · string

Custom reverse proxy for Anthropic API

Anthropic via Google Vertex AI

ANTHROPIC_USE_VERTEX · boolean

Use Anthropic models through Google Vertex AI instead of direct API

ANTHROPIC_VERTEX_REGION · string · default: "us-east5"

Google Vertex AI region for Anthropic models

endpoints:
  anthropic:
    streamRate: 20
    titleModel: claude-3.5-haiku
    vertex:
      region: "us-east5"
      serviceKeyFile: "/path/to/service-account.json"
      models:
        claude-opus-4.5:
          deploymentName: claude-opus-4-5@20251101
        claude-sonnet-4:
          deploymentName: claude-sonnet-4-20250514

Google (Gemini)

GOOGLE_KEY · string · required

Google API key (for Gemini API/AI Studio)

GOOGLE_KEY=user_provided

GOOGLE_MODELS · string

Comma-separated list of Google models

Gemini API

GOOGLE_MODELS=gemini-3.1-pro-preview,gemini-2.5-pro,gemini-2.5-flash,gemini-2.0-flash

GOOGLE_REVERSE_PROXY · string

Custom reverse proxy for Google API

GOOGLE_AUTH_HEADER · boolean

Pass API key in Authorization header instead of X-goog-api-key

Some reverse proxies don’t support the X-goog-api-key header

Vertex AI Configuration

GOOGLE_SERVICE_KEY_FILE · string

Path to Google Cloud service account JSON file

GOOGLE_SERVICE_KEY_FILE=/path/to/service-account.json

GOOGLE_LOC · string · default: "us-central1"

Google Cloud region for Vertex AI

GOOGLE_CLOUD_LOCATION · string · default: "global"

Alternative region for Gemini Image Generation

Safety Settings

GOOGLE_SAFETY_* · string

Google Safety Settings (apply to both Vertex AI and Gemini API)

Settings

GOOGLE_SAFETY_SEXUALLY_EXPLICIT=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_HATE_SPEECH=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_HARASSMENT=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_DANGEROUS_CONTENT=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_CIVIC_INTEGRITY=BLOCK_ONLY_HIGH

Options: BLOCK_NONE, BLOCK_ONLY_HIGH, BLOCK_MEDIUM_AND_ABOVE, BLOCK_LOW_AND_ABOVE

For Vertex AI, BLOCK_NONE requires allowlist access or monthly invoiced billing

AWS Bedrock

BEDROCK_AWS_DEFAULT_REGION · string · required

AWS region for Bedrock

BEDROCK_AWS_DEFAULT_REGION=us-east-1

BEDROCK_AWS_ACCESS_KEY_ID · string

AWS access key ID

BEDROCK_AWS_SECRET_ACCESS_KEY · string

AWS secret access key

BEDROCK_AWS_SESSION_TOKEN · string

AWS session token (for temporary credentials)

BEDROCK_AWS_MODELS · string

Comma-separated list of Bedrock model IDs

BEDROCK_AWS_MODELS=anthropic.claude-sonnet-4-6,anthropic.claude-opus-4-6-v1,meta.llama3-1-8b-instruct-v1:0

If omitted, all known supported models are included

Bedrock Configuration (YAML)

endpoints:
  bedrock:
    models:
      - "anthropic.claude-3-7-sonnet-20250219-v1:0"
      - "anthropic.claude-3-5-sonnet-20241022-v2:0"
    
    inferenceProfiles:
      "us.anthropic.claude-sonnet-4-20250514-v1:0": "${BEDROCK_INFERENCE_PROFILE_CLAUDE_SONNET}"
      "anthropic.claude-3-7-sonnet-20250219-v1:0": "arn:aws:bedrock:us-west-2:123456789012:application-inference-profile/abc123"
    
    guardrailConfig:
      guardrailIdentifier: "your-guardrail-id"
      guardrailVersion: "1"
      trace: "enabled"

Azure OpenAI

Azure environment variables are DEPRECATED. Use librechat.yaml configuration instead.

See the Custom Endpoints documentation for Azure OpenAI configuration.

Assistants API

ASSISTANTS_API_KEY · string · required

API key for OpenAI Assistants

ASSISTANTS_API_KEY=user_provided

ASSISTANTS_BASE_URL · string

Custom base URL for Assistants API

ASSISTANTS_MODELS · string

Comma-separated list of models available for Assistants

ASSISTANTS_MODELS=gpt-4o,gpt-4o-mini,gpt-4-turbo-preview

Assistants Configuration (YAML)

endpoints:
  assistants:
    disableBuilder: false
    pollIntervalMs: 3000
    timeoutMs: 180000
    supportedIds: ["asst_supportedAssistantId1", "asst_supportedAssistantId2"]
    retrievalModels: ["gpt-4-turbo-preview"]
    capabilities: ["code_interpreter", "retrieval", "actions", "tools", "image_vision"]

Known AI Providers

These providers can be configured via environment variables and used with custom endpoints:

Provider Keys

ANYSCALE_API_KEY=
APIPIE_API_KEY=
COHERE_API_KEY=
DEEPSEEK_API_KEY=
DATABRICKS_API_KEY=
FIREWORKS_API_KEY=
GROQ_API_KEY=
HUGGINGFACE_TOKEN=
MISTRAL_API_KEY=
OPENROUTER_KEY=
PERPLEXITY_API_KEY=
SHUTTLEAI_API_KEY=
TOGETHERAI_API_KEY=
UNIFY_API_KEY=
XAI_API_KEY=

See Custom Endpoints for configuration examples.

Agents Endpoint

endpoints:
  agents:
    recursionLimit: 50
    maxRecursionLimit: 100
    disableBuilder: false
    maxCitations: 30
    maxCitationsPerFile: 7
    minRelevanceScore: 0.45
    capabilities: ["deferred_tools", "execute_code", "file_search", "actions", "tools"]

agents.recursionLimit · number · default: "25"

Default recursion depth for agents

agents.maxRecursionLimit · number · default: "25"

Maximum recursion depth for agents

agents.maxCitations · number · default: "30"

Maximum total citations in agent responses

agents.minRelevanceScore · number · default: "0.45"

Minimum relevance score for sources (0.0-1.0)

Set to 0.0 to show all sources without filtering