Market data
Polymarket Gamma API
It is the core market-data layer for finding live AI prediction markets and ranking them by volume.
Cadence: Daily static snapshot, with room for more frequent refreshes.
Used for
- AI market discovery
- Market title and slug normalization
- One-month volume and liquidity context
- Current yes-probability extraction where outcome prices are available
Limitations
- Some event slugs differ from seeded Odsage slugs and require normalization.
- Resolution wording still needs human review before an article treats a market as high confidence.
Market data
Polymarket CLOB API
Order-book context can separate a real market move from a thin-liquidity price print.
Cadence: Planned for intraday pricing and order-book depth checks.
Used for
- More reliable yes/no price checks
- Spread and depth analysis
- Liquidity warnings on calculator pages
Limitations
- Not yet part of the committed static snapshot.
- Must stay server-side so no user-facing bundle exposes integration details.
It provides a normalized prediction-market search layer that can cross-check Polymarket market prices without calling third-party APIs from public pages.
Cadence: Daily static snapshot during the AI-market intelligence refresh.
Used for
- External Polymarket price cross-checks
- Bid, ask, last price, and spread context
- Market URL and match-confidence enrichment
- Future multi-source prediction-market comparisons
Limitations
- Prices are snapshots and can move before the next daily refresh.
- Search match confidence is a discovery signal, not a guarantee that the result matches every resolution detail.
Keyword data
DataForSEO Google Ads keyword data
It tells Odsage which AI-odds topics have search demand before larger sites create dedicated pages.
Cadence: Daily during the AI-market intelligence refresh.
Used for
- Low-competition article discovery
- Search-volume and CPC labels
- Keyword-opportunity scoring
- Editorial queue prioritization
Limitations
- Zero public Google Ads volume does not mean zero search intent.
- Keyword suggestions need relevance filtering to remove unrelated literal-name matches.
News
DataForSEO Google News SERP
It keeps Odsage focused on fresh market-moving AI news without republishing publisher articles.
Cadence: Daily US-morning market brief.
Used for
- AI model news discovery
- Company/entity detection
- Market-angle generation
- Daily impact scoring
Limitations
- News snippets can be noisy and must be summarized into original Odsage market context.
- Publisher dates and index freshness should be interpreted carefully.
Social sentiment
xAI Grok API
Grok can analyze fast-moving AI discourse and summarize sentiment pressure as a layer beside price, liquidity, and news impact.
Cadence: Prepared for server-side sentiment scoring after usage limits are confirmed.
Used for
- AI-market sentiment classification
- Driver-term extraction from social and news text
- Market-impact summaries that avoid trade recommendations
Limitations
- An xAI key analyzes supplied text; it does not by itself provide full X/Twitter post ingestion.
- Model output must stay server-side and should be cached before public rendering.
Social sentiment
last30days social research skill
It gives Odsage a compact view of what high-engagement social accounts are saying about AI model prediction markets over the last 30 days.
Cadence: Lightweight X-only pass during the static intelligence refresh when the local skill is available.
Used for
- Recent X/social discussion pulse
- Top voice and engagement extraction
- Supporting context beside the primary Odsage sentiment score
Limitations
- It is a supporting signal, not the primary sentiment engine.
- The local CLI can fail if optional platform dependencies are missing, so Odsage keeps the previous snapshot instead of blocking refreshes.
Benchmark
Artificial Analysis
It gives a structured way to compare model capability, latency, and pricing against market odds.
Cadence: Manual/contextual until an allowed adapter is added.
Used for
- Model quality context
- Speed and price comparisons
- Best-model market evidence
Limitations
- Odsage should cite and link rather than scrape restricted data.
- Benchmark leadership can change faster than article update cadence.
Preference rankings are directly relevant to markets asking which lab has the strongest model.
Cadence: Manual/contextual until an allowed adapter is added.
Used for
- Human preference leaderboard context
- Best-model market evidence
- Model-family comparison pages
Limitations
- Leaderboard results are not identical to Polymarket resolution rules.
- Preference data should be combined with technical benchmarks and release evidence.
It adds a harder-to-game benchmark layer for frontier model comparisons.
Cadence: Manual/contextual until an allowed adapter is added.
Used for
- Contamination-resistant benchmark context
- Reasoning and coding signal checks
- Release-market evidence
Limitations
- A benchmark score is evidence, not a market-resolution guarantee.
- Coverage depends on which model versions are included at a given time.
Search performance
Google Search Console
It closes the SEO feedback loop from keyword research to actual Google visibility.
Cadence: After production indexing and verified property access.
Used for
- Article impression tracking
- Query-level content expansion
- Low-CTR page improvement
Limitations
- Requires verified production property access.
- Data is delayed and should not drive same-day market decisions.
It helps search engines discover new market and article URLs after publication.
Cadence: After successful production deploys with new URLs.
Used for
- Fresh URL discovery
- Post-deploy sitemap submission workflow
- Faster indexing notifications for supported engines
Limitations
- It is a discovery signal, not a ranking guarantee.
- Google indexing still depends on Google Search Console and crawl behavior.