Skip to main content

Configuration

Ideon merges configuration from multiple sources and validates the result before execution.

Precedence Rules​

Lowest to highest priority:

  1. Saved settings file
  2. Job file settings
  3. Environment variables
  4. Direct CLI arguments (--style, --intent, --primary, --secondary, idea input)

Secret precedence:

  • IDEON_OPENROUTER_API_KEY and IDEON_REPLICATE_API_TOKEN from environment variables override keychain-stored secrets.
  • If env vars are not set, Ideon attempts to read keychain values saved through ideon settings.
  • Keychain support (keytar) is loaded lazily at runtime when secret reads/writes are needed.
  • If keychain access fails (for example D-Bus is unavailable in a container), Ideon falls back to environment variables for secret resolution.
  • Set IDEON_DISABLE_KEYTAR=true to skip keychain access entirely in container or CI environments.

Per-field merge behavior:

  • modelSettings merges by key (temperature, maxTokens, topP) across sources.
  • contentTargets is replaced as a full array when provided by a higher-priority source.
  • Scalar settings (for example model, style, intent, targetLength, markdownOutputDir) are replaced by the highest-priority source.

Settings Schema​

Core settings include:

  • model: LLM model identifier
  • modelSettings.temperature: 0..2
  • modelSettings.maxTokens: positive integer
  • modelSettings.topP: 0..1
  • modelRequestTimeoutMs: positive integer request timeout in milliseconds (default 90000)
  • t2i.modelId: selected text-to-image model
  • t2i.inputOverrides: model-specific user overrides
  • notifications.enabled: toggles OS notifications for write lifecycle updates
  • markdownOutputDir
  • assetOutputDir
  • contentTargets: array of output targets with per-type counts
  • style: run-level writing style
  • intent: run-level content intent
  • targetLength: run-level target length in words (positive integer). Aliases are accepted as input: small=500, medium=900, large=1400.

contentTargets entries:

  • contentType: one of article, blog-post, linkedin-post, newsletter, press-release, reddit-post, science-paper, x-post, x-thread
  • role: primary or secondary
  • count: positive integer

Rules:

  • Exactly one contentTargets entry must have role primary.
  • Primary count must be 1.
  • Secondary entries are optional and can use counts greater than 1.

Style values:

  • academic
  • analytical
  • authoritative
  • conversational
  • empathetic
  • friendly
  • journalistic
  • minimalist
  • persuasive
  • playful
  • professional
  • storytelling
  • technical

Intent values:

  • announcement
  • case-study
  • cornerstone
  • counterargument
  • critique-review
  • deep-dive-analysis
  • how-to-guide
  • interview-q-and-a
  • listicle
  • opinion-piece
  • personal-essay
  • roundup-curation
  • tutorial

Defaults:

  • contentTargets: [ { "contentType": "article", "role": "primary", "count": 1 } ]
  • style: professional
  • intent: tutorial
  • targetLength: 900

Target length aliases:

  • small: 500 words
  • medium: 900 words (default)
  • large: 1400 words

Saved Settings Location​

Saved via OS config path (using env-paths), typically:

  • macOS: ~/.ideon/settings.json

To edit saved settings, run ideon settings again. The wizard is the supported way to update values and stored credentials.

Example Environment Override​

IDEON_MODEL=openai/gpt-4.1-mini \
IDEON_TEMPERATURE=0.6 \
IDEON_MAX_TOKENS=2400 \
IDEON_STYLE=technical \
IDEON_INTENT=tutorial \
IDEON_TARGET_LENGTH=1200 \
ideon write "An idea"

Note: content target arrays are not currently configurable through environment variables. Use CLI --primary/--secondary flags or job-file settings.contentTargets.

See Environment Variables for full list.