Settings Reference All WikiMind configuration is managed via environment variables, loaded by Pydantic BaseSettings . A .env file at the project root is loaded automatically.
WikiMind uses the WIKIMIND_ prefix for all settings. Nested settings use the __ delimiter:
WIKIMIND_LLM__OPENAI__ENABLED = true
# Sets: settings.llm.openai.enabled = True
LLM API Keys Set at least one API key. The provider auto-enables when a key is detected.
Variable Description ANTHROPIC_API_KEYAnthropic API key OPENAI_API_KEYOpenAI API key GOOGLE_API_KEYGoogle Gemini API key
Tip
API keys work both with and without the WIKIMIND_ prefix (e.g., both ANTHROPIC_API_KEY and WIKIMIND_ANTHROPIC_API_KEY are accepted). Keys are also read from the OS keychain via keyring.
LLM Provider Configuration Variable Default Description WIKIMIND_LLM__DEFAULT_PROVIDERanthropicDefault LLM provider (anthropic, openai, google, ollama, mock) WIKIMIND_LLM__FALLBACK_ENABLEDtrueFall back to other providers when the default fails WIKIMIND_LLM__MONTHLY_BUDGET_USD50.0Monthly spending ceiling across all providers (USD) WIKIMIND_LLM__BUDGET_WARNING_PCT0.8Budget warning threshold (fraction 0.0-1.0) WIKIMIND_LLM__BUDGET_CHECK_CACHE_SECONDS60How often the router re-queries cost logs
Per-Provider Settings Each provider has model and enabled settings:
Variable Default Description WIKIMIND_LLM__ANTHROPIC__MODELclaude-sonnet-4-5Anthropic model name WIKIMIND_LLM__ANTHROPIC__ENABLEDauto Enabled when API key detected WIKIMIND_LLM__OPENAI__MODELgpt-4oOpenAI model name WIKIMIND_LLM__OPENAI__ENABLEDauto Enabled when API key detected WIKIMIND_LLM__GOOGLE__MODELgemini-2.0-flashGoogle Gemini model name WIKIMIND_LLM__GOOGLE__ENABLEDauto Enabled when API key detected WIKIMIND_LLM__OLLAMA__MODELllama3.2Ollama model name WIKIMIND_LLM__OLLAMA__ENABLEDfalseMust be explicitly enabled (no API key) WIKIMIND_LLM__OLLAMA_BASE_URLhttp://localhost:11434Ollama server URL WIKIMIND_LLM__MOCK__ENABLEDfalseEnable mock provider (CI/testing only)
Server Variable Default Description WIKIMIND_SERVER__HOST127.0.0.1Server bind address WIKIMIND_SERVER__PORT7842Server port WIKIMIND_GATEWAY_PORT7842Gateway port (used by lifespan logging)
Database Variable Default Description WIKIMIND_DATABASE_URLsqlite+aiosqlite:///~/.wikimind/db/wikimind.dbDatabase connection URL WIKIMIND_DATABASE__ECHOfalseVerbose SQL query logging DATABASE_URL-- Fallback for managed Postgres (Fly.io, Railway, etc.)
The DATABASE_URL fallback auto-rewrites postgres:// and postgresql:// to postgresql+asyncpg:// for SQLAlchemy async compatibility. On Fly.io, sslmode is converted for asyncpg.
Data Directory Variable Default Description WIKIMIND_DATA_DIR~/.wikimindBase data directory WIKIMIND_STORAGE_BACKENDlocalStorage backend (local or r2)
Job Queue Variable Default Description WIKIMIND_REDIS_URL-- Redis URL for ARQ job queue REDIS_URL-- Fallback (unprefixed) Redis URL
When unset, compilations run in-process (single-user dev mode). No Redis needed.
Variable Default Description WIKIMIND_DOCLING_SERVE_URLhttp://localhost:5001docling-serve sidecar URL WIKIMIND_DOCLING_BATCH_PAGES10Pages per Docling batch
Vision Processing Variable Default Description WIKIMIND_VISION_ENABLEDtrueEnable vision-enhanced slide deck ingestion WIKIMIND_VISION_TEXT_THRESHOLD300Characters below which a page is treated as image-heavy WIKIMIND_VISION_DPI150DPI for rendering pages as images WIKIMIND_VISION_MAX_PAGES_PER_BATCH20Max pages per vision batch
Variable Default Description WIKIMIND_IMAGE_EXTRACTION_ENABLEDtrueExtract figures/tables from PDFs WIKIMIND_IMAGE_MAX_PER_PDF30Maximum images to extract per PDF WIKIMIND_IMAGE_BASE_URL/imagesBase URL for serving extracted images
Q&A Agent Variable Default Description WIKIMIND_QA__MAX_PRIOR_TURNS_IN_CONTEXT5Prior conversation turns included in context WIKIMIND_QA__PRIOR_ANSWER_TRUNCATE_CHARS500Max characters per prior answer in context WIKIMIND_QA__CONVERSATION_TITLE_MAX_CHARS120Max characters for auto-generated conversation titles
Concept Taxonomy Variable Default Description WIKIMIND_TAXONOMY__REBUILD_THRESHOLD5New concepts before auto-triggering taxonomy rebuild WIKIMIND_TAXONOMY__MAX_HIERARCHY_DEPTH3Maximum depth of concept hierarchy WIKIMIND_TAXONOMY__CONCEPT_PAGE_MIN_SOURCES2Minimum sources before generating a concept page
Wiki Linter Variable Default Description WIKIMIND_LINTER__ENABLE_ORPHAN_DETECTIONtrueEnable orphan article detection WIKIMIND_LINTER__MAX_CONCEPTS_PER_RUN25Max concepts to check per lint run WIKIMIND_LINTER__MAX_CONTRADICTION_PAIRS_PER_CONCEPT10Max article pairs per concept for contradiction detection WIKIMIND_LINTER__CONTRADICTION_LLM_MAX_TOKENS1024Max tokens for contradiction LLM calls WIKIMIND_LINTER__CONTRADICTION_LLM_TEMPERATURE0.2Temperature for contradiction detection WIKIMIND_LINTER__ENABLE_PAIR_CACHEtrueSkip re-evaluating unchanged article pairs WIKIMIND_LINTER__CONTRADICTION_BATCH_ENABLEDtrueBatch multiple pairs into single LLM calls WIKIMIND_LINTER__CONTRADICTION_BATCH_SIZE10Pairs per batch WIKIMIND_LINTER__MAX_CONCEPT_CONCURRENCY5Concurrent concept checks
Embedding / Semantic Search Variable Default Description WIKIMIND_EMBEDDING__MODEL_NAMEall-MiniLM-L6-v2Sentence-transformers model WIKIMIND_EMBEDDING__CHUNK_SIZE_TOKENS500Chunk size for embedding WIKIMIND_EMBEDDING__CHUNK_OVERLAP_TOKENS50Overlap between chunks WIKIMIND_EMBEDDING__MIN_SIMILARITY_SCORE0.65Minimum similarity for search results
Authentication See the Authentication page for detailed setup.
Variable Default Description WIKIMIND_AUTH__ENABLEDfalseEnable multi-user authentication WIKIMIND_AUTH__JWT_SECRET_KEY-- Secret key for JWT signing (required when auth enabled) WIKIMIND_AUTH__JWT_ALGORITHMHS256JWT algorithm WIKIMIND_AUTH__JWT_EXPIRY_MINUTES1440JWT expiry (24 hours) WIKIMIND_AUTH__COOKIE_NAMEwikimind_sessionSession cookie name WIKIMIND_AUTH__COOKIE_SECUREtrueRequire HTTPS for cookies WIKIMIND_AUTH__COOKIE_DOMAIN-- Cookie domain (None = current host)
Production Deployment Variable Default Description POSTGRES_USERwikimindPostgreSQL username (docker-compose.prod.yml) POSTGRES_PASSWORD-- PostgreSQL password (required) POSTGRES_DBwikimindPostgreSQL database name WIKIMIND_PORT7842Host port for the gateway WIKIMIND_IMAGEwikimind:latestDocker image tag WEB_CONCURRENCYauto Gunicorn worker count (auto-tunes to CPU) LOG_LEVELINFOLogging level
April 21, 2026 April 21, 2026