.env.example

# =============================================================================
# ContextSeek Configuration Template
# =============================================================================
# Copy this file to .env and adjust values for your environment.
#
# Optional: place a shared config at examples/configs/.env. Resolution order:
# cwd .env → repo root .env → examples/configs/.env → environment variables.
#
# Each block uses an env prefix + field name (STORAGE_ + BACKEND →
# STORAGE_BACKEND). Names are matched case-insensitively.
#
# Required for minimal use: nothing (defaults are in-memory, no models).
# Optional: file storage, LangChain embedding/LLM, retrieval tuning, evolution,
# observability, lifecycle.
# =============================================================================

# =============================================================================
# 1. Storage
# =============================================================================
# Backend: "seekdb" (embedded, persistent — default) | "memory" (ephemeral) |
#          "file" (persistent) | "oceanbase" (requires OB_* below)
STORAGE_BACKEND=seekdb
# Root directory when backend is "file"
STORAGE_PATH=.contextseek/store
# URI scheme for scope resolution
# STORAGE_URI_SCHEME=contextseek://
# Optional cold tier: "file" or "memory" (leave empty to disable)
# STORAGE_COLD_BACKEND=
# STORAGE_COLD_PATH=.contextseek/cold

# =============================================================================
# 1.0 SeekDB Storage (used when STORAGE_BACKEND=seekdb)
# =============================================================================
# Embedded mode by default (local .db directory); keep STORAGE_BACKEND=seekdb.
# For seekdb server mode, set STORAGE_BACKEND=oceanbase instead.
# SEEKDB_PATH=~/.contextseek/seekdb.db
# SEEKDB_HOST=                       # empty = embedded; set host for remote server mode
# SEEKDB_PORT=2881
# SEEKDB_DATABASE=contextseek

# =============================================================================
# 1.1 OceanBase Storage (required when STORAGE_BACKEND=oceanbase)
# =============================================================================
# Set STORAGE_BACKEND=oceanbase and fill in the OB_* params below.
# EMBEDDING_DIMS must also be set (section 2) — OceanBase needs the vector size
# to create its HNSW index at table creation time.
#
# OB_HOST=127.0.0.1
# OB_PORT=2881
# OB_USER=root@test
# OB_PASSWORD=
# OB_DB_NAME=contextseek
# OB_TABLE_NAME=contextseek_items

# =============================================================================
# 2. Embedding
# =============================================================================
# Provider: "none" | "openai" | "dashscope" | "ollama" | "huggingface" | "langchain"
EMBEDDING_PROVIDER=none
# EMBEDDING_MODEL=text-embedding-3-small
# Optional: inferred from known providers when omitted
# EMBEDDING_DIMS=1536
# Optional base URL for compatible providers (OpenAI-compatible, Ollama, etc.)
# EMBEDDING_BASE_URL=

## Typical provider snippets (uncomment one block as needed)
# OpenAI:
#   EMBEDDING_PROVIDER=openai
#   EMBEDDING_MODEL=text-embedding-3-small
#   OPENAI_API_KEY=sk-...
#   EMBEDDING_BASE_URL=https://api.openai.com/v1
#
# DashScope:
#   EMBEDDING_PROVIDER=dashscope
#   EMBEDDING_MODEL=text-embedding-v3
#   DASHSCOPE_API_KEY=sk-...
#
# Ollama (local):
#   EMBEDDING_PROVIDER=ollama
#   EMBEDDING_MODEL=nomic-embed-text
#   EMBEDDING_BASE_URL=http://127.0.0.1:11434
#
# HuggingFace (local):
#   EMBEDDING_PROVIDER=huggingface
#   EMBEDDING_MODEL=BAAI/bge-small-zh-v1.5
#
# Advanced custom LangChain class:
#   EMBEDDING_PROVIDER=langchain
#   EMBEDDING_CLASS_PATH=your_package.YourEmbeddings
#   EMBEDDING_DIMS=1536

# =============================================================================
# 3. LLM (rerank / extract / summarize)
# =============================================================================
# Provider: "none" | "openai" | "dashscope" | "ollama" | "langchain"
LLM_PROVIDER=none
# LLM_MODEL=gpt-4o-mini
# Optional base URL for compatible providers (OpenAI-compatible, Ollama, etc.)
# LLM_BASE_URL=

## Typical provider snippets
# OpenAI:
#   LLM_PROVIDER=openai
#   LLM_MODEL=gpt-4o-mini
#   OPENAI_API_KEY=sk-...
#   LLM_BASE_URL=https://api.openai.com/v1
#
# DashScope:
#   LLM_PROVIDER=dashscope
#   LLM_MODEL=qwen-plus
#   DASHSCOPE_API_KEY=sk-...
#
# Ollama:
#   LLM_PROVIDER=ollama
#   LLM_MODEL=qwen2.5:7b
#   LLM_BASE_URL=http://127.0.0.1:11434
#
# Advanced custom LangChain class:
#   LLM_PROVIDER=langchain
#   LLM_CLASS_PATH=your_package.YourChatModel

# =============================================================================
# 4. Summarizer (drives L2 abstract / L1 overview generation in ContextSeek.add)
# =============================================================================
# Provider: "none" (disabled) | "llm" (uses LLM_*)
# SUMMARIZER_PROVIDER=none
# SUMMARIZER_L2_MAX_CHARS=100        # L2 abstract char budget
# SUMMARIZER_L1_MAX_CHARS=2000       # L1 overview char budget

# =============================================================================
# 5. Retrieval
# =============================================================================
RETRIEVAL_DEFAULT_K=20
# RETRIEVAL_DEFAULT_BUDGET=20
# RETRIEVAL_RECALL_ROUTES=["phrase","terms"]
# RETRIEVAL_RECALL_ROUTES=["phrase","terms","vector"]
# RETRIEVAL_RERANKER_MODE=heuristic  # heuristic | llm
# RETRIEVAL_LLM_RERANK_TOP_N=20
# RETRIEVAL_VECTOR_WEIGHT=0.7
# RETRIEVAL_FTS_WEIGHT=0.3
# RETRIEVAL_MAX_CONTENT_CHARS=1200

# =============================================================================
# 6. Evolution (merge / extract / distill)
# =============================================================================
EVOLUTION_ENABLED=false
# EVOLUTION_SEMANTIC_MERGE_THRESHOLD=0.72
# EVOLUTION_MIN_CLUSTER_SIZE=3
# EVOLUTION_DECAY_HALF_LIFE_DAYS=7.0
# EVOLUTION_EPHEMERAL_TTL_SECONDS=3600.0
# EVOLUTION_LLM_MERGE_ENABLED=false
# EVOLUTION_LLM_CONFLICT_CHECK_ENABLED=false
# EVOLUTION_LLM_STAGE_INFER_ENABLED=false
# EVOLUTION_LLM_DISTILL_ENABLED=false
# EVOLUTION_LLM_FEEDBACK_ENABLED=false

# =============================================================================
# 6.1 Dream (optional LLM in consolidation/divergence)
# =============================================================================
# DREAM_LLM_ENABLED=false

# =============================================================================
# 7. Prompt Templates (optional overrides for LLM prompts)
# =============================================================================
# Values use Python format placeholders, e.g. {query}, {content}, {items}.
# For JSON examples in templates, escape braces as {{ and }}.
#
# PROMPT_SUMMARIZER_ABSTRACT_TEMPLATE=In at most {char_budget} chars, summarize:\n\n{content}
# PROMPT_SUMMARIZER_OVERVIEW_TEMPLATE=In at most {char_budget} chars, provide structured overview:\n\n{content}
# PROMPT_RETRIEVAL_RELEVANCE_TEMPLATE=Score 0..1 and return JSON {{\"score\": <float>}}.\nQuery: {query}\nPassage: {content}
# PROMPT_CONFLICT_JUDGE_TEMPLATE=Classify as near_duplicate/contradiction/none and return JSON {{\"label\":\"...\"}}.\nA: {new_text}\nB: {existing_text}
# PROMPT_STAGE_CLASSIFIER_TEMPLATE=Infer stage and return JSON {{\"stage\":\"...\"}}.\nsource_type: {source_type}\ndefault_stage: {default_stage}\ncontent: {content_text}
# PROMPT_FEEDBACK_TAG_TEMPLATE=Return JSON {{\"tag\":\"...\"}} for stage {stage} reason {reason}
# PROMPT_MERGE_SYNTHESIS_TEMPLATE=Synthesize one statement.\n{items}
# PROMPT_DISTILL_CANDIDATE_TEMPLATE=Return JSON {{\"distill\": true|false}}.\ncontent: {content_text}\ntags: {tags}
# PROMPT_DISTILL_RENDER_TEMPLATE=Return JSON with name/description/body.\ncontent: {content_text}\ntags: {tags}
# PROMPT_DREAM_CONSOLIDATION_TEMPLATE=Summarize shared pattern.\n{items}
# PROMPT_DREAM_DIVERGENCE_TEMPLATE=A: {a_text}\nB: {b_text}\nGenerate concise hypothesis.

# =============================================================================
# 8. Security
# =============================================================================
# SECURITY_ACL_ENABLED=true
# SECURITY_REDACT_SENSITIVE=false

# =============================================================================
# 9. Observability
# =============================================================================
OBSERVABILITY_AUDIT_ENABLED=false
# OBSERVABILITY_AUDIT_PATH=.contextseek/audit.jsonl
# OBSERVABILITY_METRICS_ENABLED=false
# OBSERVABILITY_METRICS_PATH=.contextseek/metrics.prom

# LangSmith tracing (https://smith.langchain.com)
# LANGSMITH_TRACING=false
# LANGSMITH_API_KEY=
# LANGSMITH_PROJECT=demo
# LANGSMITH_ENDPOINT=https://api.smith.langchain.com

# =============================================================================
# 10. Lifecycle
# =============================================================================
# LIFECYCLE_INTERVAL_SECONDS=3600
# LIFECYCLE_AUTO_COMPACT=true
# LIFECYCLE_COMPACT_MIN_ITEMS=5

# =============================================================================
# 11. GIS（need STORAGE_BACKEND=oceanbase，OceanBase >= 4.2.2 or seekdb）
# =============================================================================
# GEO_ENABLED=false
# GEO_GEO_TABLE_NAME=contextseek_geo
# GEO_SRID=4326