Configuration reference

July 31, 2026 · View on GitHub

All plugins in this package share the same OpenAI client configuration. Each plugin reads the keys it needs from its own config dictionary.

Common keys

KeyTypeDefaultDescription
api_urlstrhttps://api.openai.com/v1Base URL of the OpenAI-compatible server (the /v1 root). /chat/completions is appended automatically.
keystr""API key sent as Authorization: Bearer <key>. Optional for local servers that do not require authentication.
modelstrgpt-4o-miniModel identifier passed to the server.
system_promptstr(plugin specific)System instruction prepended to the conversation.

Generation parameters

These are forwarded to the Chat Completions request body:

KeyTypeDefaultDescription
max_tokensint300Maximum number of tokens to generate.
temperaturefloat0.5Sampling temperature (0 to 2).
top_pfloat0.2Nucleus sampling probability mass.
frequency_penaltyfloat0Penalize repeated tokens (−2 to 2).
presence_penaltyfloat0Encourage new topics (−2 to 2).
stop_tokenstr / listNoneStop sequence(s).

Chat engine specific keys

KeyTypeDefaultDescription
allow_system_promptsboolfalseIf false, system messages supplied by the caller are stripped and replaced by system_prompt. If true, a caller-supplied system message is merged with system_prompt.

Summarizer specific keys

KeyTypeDefaultDescription
prompt_templatestrbuilt-inTemplate containing {content}, filled with the document to summarize.

Example

{
  "api_url": "https://api.openai.com/v1",
  "key": "sk-xxxx",
  "model": "gpt-4o-mini",
  "system_prompt": "You are a concise assistant.",
  "max_tokens": 300,
  "temperature": 0.5
}

Home · Available plugins →