PredictionToolkitWhat is Eversight and Signals APIs?
Eversight accepts a precise question as an asynchronous job. Poll until COMPLETED before reading result.prediction on a 0–1 scale. Signals accepts supported market URLs, slugs, condition IDs or questions and returns weighted material for research.
Use as_of to restrict historical retrieval, and distinguish published_at from capture time. Numinous-1 is a specialized forecasting model with an OpenAI-compatible endpoint and structured output, not a general assistant. Its documented context window is 16,384 tokens; inspect finish_reason and parsing if an answer is cut short.
Who it is for
- Developers building research assistants.
- Analysts comparing forecasts with market prices.
- Teams needing attributable evidence.
Research coverage and methods
- Forecast jobs
Submit questions, poll states and read completed probabilities.
- Market-aware signals
Use URLs, slugs, condition IDs or free text.
- Attribution
Inspect material behind weighted signals.
- Historical retrieval
Use as_of to preserve an earlier information boundary.
- News impact
Read supported scored signals and their direction.
- Related markets
Explore causal relationships with associated evidence.
- Numinous-1
Use the specialized forecasting model through an OpenAI-compatible endpoint with structured-output support.
Research tasks
Compare a forecast
Read exact resolution terms, submit a job and wait for completion. Compare probability and quote while retaining evidence.
Build a research panel
Retrieve Signals, display linked material and separate publication and snapshot times.
Replay past research
Set as_of before retrieval so later articles stay outside the earlier decision window.
Pricing and total cost
Metered access uses supported API-key or x402 payment paths. Forecast jobs, signals and model inference are separate operations; check their rate schedule before batching.
Research inputs and access
- Choose Eversight for jobs or Signals for evidence.
- Configure the documented API-key or x402 route.
- Start with a precise question or supported identifier.
- Poll forecasts until COMPLETED before reading result.prediction.
- Store sources and the query time boundary.
- Handle truncation and structured parsing for model inference.
Useful links
What to check
- Numinous-1 specializes in forecasting rather than general assistance.
- Use either event_id or condition_id when mutually exclusive.
- Publication and capture timestamps can differ.
- The documented context window is 16,384 tokens; inspect finish_reason for truncated output.
A closer look
Ratings & reviews
User reviews 0
The main score switches to user ratings after 5 published reviews.
A specific example helps the next person choose the right tool.
Related tools
Browse category
Request forecasts and market evidence from an agent while tracking separately metered API usage.
- Language Python
Request an AI forecast from a link, question or image, then compare market odds and news context.
- Access Web
Paste a Polymarket or Kalshi contract link to start a focused AI research task through Valyu.
- Access Web
Frequently asked questions about Eversight and Signals APIs
Where do I retrieve supporting articles?
Use Signals retrieval endpoints.
How is probability returned?
Completed jobs return result.prediction from 0 to 1.
Can I exclude later information?
Use documented as_of retrieval.
Do these APIs execute trades?
No. They supply research and probabilities; execution requires another integration.