Skip to main content

AI in Signailist

Overview​

AI is used in four primary workflows:

  1. Relevance scoring of news against workspace themes.
  2. Text generation for digests and publication drafts.
  3. Template generation from Telegram sources through style extraction.
  4. Image generation for digest covers and related visuals.

All AI features currently run through the OpenAI API.

1. Theme-based relevance scoring​

How scoring works​

Each theme is defined as an AI-first interest profile:

  • What to look for: a plain-language description of relevant content.
  • What to exclude: a plain-language description of irrelevant content.

Each incoming news item is evaluated against each active theme with:

  • Relevance: whether the item matches the theme.
  • Score: a value between 0 and 1.
  • Confidence: how sure the model is.
  • Reason: a short explanation of the decision.

Scoring settings​

At workspace level, teams can configure:

  • The OpenAI model used for scoring.
  • Request-per-minute limits.
  • Processing concurrency.

If an item has already been scored for the same theme, repeated scoring is skipped automatically.

2. Digest text generation​

How generation works​

  1. A user selects digest candidates and a template.
  2. AI generates the digest text from the chosen news items.
  3. The user reviews, edits, and approves the result.
  4. The final text is split into Telegram-sized chunks when needed.

See Digests for the full workflow.

3. Template generation from a source​

How template generation works​

Template generation analyzes an existing Telegram source and creates a reusable writing template from that source's tone and structure.

  1. The user selects a source and a sampling window.
  2. AI analyzes tone, structure, and writing patterns.
  3. A template is produced with style settings.
  4. Optional test generations can be produced to validate the output quality.

Template generation​

Template generation analyzes source posts and creates a style profile with optional test generations for quality validation.

4. Image generation​

How image generation works​

For a digest, Signailist can generate a cover image:

  1. AI builds an image prompt from the digest content and the template settings.
  2. The image is generated through OpenAI image generation.

Image settings​

Each digest template can define:

  • Whether image generation is enabled.
  • A base image prompt.
  • Preferred size or aspect.
  • Reference images.

Limitations​

  • Without an OpenAI API key, AI features are not available.
  • Models can be configured at workspace level for scoring and at template level for generation.
  • Image availability depends on the active generation and storage flow configured in the product.