mirror of
https://github.com/open-webui/docs.git
synced 2026-11-04 18:18:03 +00:00
3.8 KiB
3.8 KiB
title, sidebar_position
| title | sidebar_position |
|---|---|
| Environment variables | 3 |
Environment variables
All variables are read once at startup; restart the server to apply a change.
Core
| Variable | Default | What it does |
|---|---|---|
CPTR_DATA_DIR |
~/.cptr |
Data directory for the database, config, uploads, and logs. The Docker image sets it to /data. |
CPTR_LOG_LEVEL |
INFO |
Server log level. |
CPTR_LOG_FORMAT |
text |
text or json (structured output for log collectors). |
CPTR_AUDIT_LOG_LEVEL |
NONE |
NONE disables audit logging; METADATA records mutations; REQUEST adds redacted request bodies; REQUEST_RESPONSE adds redacted response bodies. |
CPTR_AUDIT_LOG_PATH |
<data>/logs/audit.jsonl |
Audit log file location. |
CPTR_AUDIT_LOG_ROTATION |
10 MB |
Audit log rotation size. |
CPTR_AUDIT_MAX_BODY_SIZE |
2048 |
Max bytes of request/response body captured per audit entry. |
CPTR_AUDIT_EXCLUDED_PATHS |
/api/chats,/v1/chat |
Paths excluded from audit logging. |
CPTR_LOG_UPSTREAM_REQUESTS |
false |
Log outgoing model API requests, for debugging model calls. |
CPTR_UPSTREAM_REQUEST_LOG_PATH |
<data>/logs/upstream-requests.jsonl |
Upstream request log location. |
CPTR_UPSTREAM_REQUEST_LOG_ROTATION |
50 MB |
Upstream request log rotation size. |
Chat and agent tuning
| Variable | Default | What it does |
|---|---|---|
CHAT_MAX_ITERATIONS |
2048 |
Max tool-call iterations per turn before the agent is stopped. |
ENABLE_CHAT_RECONCILE_ON_STARTUP |
true |
On startup, recover chats left stuck by a crash. |
CHAT_TOOL_MAX_CHARS |
50000 |
Max characters of a tool's output kept in context. |
CHAT_TOOL_COMMAND_MAX_CHARS |
8000 |
Max characters of command output kept in context. |
CHAT_COMPACT_TOKEN_THRESHOLD |
80000 |
Token count that triggers context compaction. |
CPTR_CLAUDE_CODE_MAX_BUFFER_SIZE |
134217728 |
Stdout buffer size (bytes) for the Claude SDK subprocess. |
CPTR_EXECUTE_TIMEOUT |
unset | Default wait in seconds for run_command before it returns while the command keeps running. |
CPTR_STREAM_CONNECT_TIMEOUT |
30 |
Seconds to wait when connecting to a model stream. |
CPTR_STREAM_READ_TIMEOUT |
300 |
Seconds to wait for the next chunk from a model stream. |
CPTR_STREAM_WRITE_TIMEOUT |
600 |
Seconds to wait when writing to a model stream. |
Scheduler
| Variable | Default | What it does |
|---|---|---|
AUTOMATION_POLL_INTERVAL |
10 |
How often (seconds) the scheduler checks for due automations. |
Network
| Variable | Default | What it does |
|---|---|---|
CPTR_CORS_ALLOWED_ORIGINS |
* |
Comma-separated list of browser origins allowed to call the API. |
Web search providers
Keys set here take precedence over keys entered in Settings. With no key at all, web search falls back to DuckDuckGo; otherwise the first configured provider in this order wins: Exa → Perplexity → Tavily → Brave → Firecrawl → SearXNG → DuckDuckGo.
| Variable | Default | What it does |
|---|---|---|
EXA_API_KEY |
unset | Exa search. |
PERPLEXITY_API_KEY |
unset | Perplexity search. |
PERPLEXITY_BASE_URL |
unset | Custom Perplexity endpoint. |
TAVILY_API_KEY |
unset | Tavily search. |
BRAVE_API_KEY |
unset | Brave search. |
FIRECRAWL_API_KEY |
unset | Firecrawl search. |
FIRECRAWL_BASE_URL |
unset | Custom Firecrawl endpoint. |
SEARXNG_BASE_URL |
unset | Self-hosted SearXNG instance. |
CHAT_COMPLETIONS_SEARCH_API_KEY |
unset | Search via an OpenAI-compatible chat-completions endpoint. |
CHAT_COMPLETIONS_SEARCH_BASE_URL |
unset | Base URL for that endpoint. |
CHAT_COMPLETIONS_SEARCH_MODEL |
unset | Model for that endpoint. |
XAI_API_KEY |
unset | Used for Grok agent detection. |