AI models & inference

Validate caller-supplied JSON against Forest bounded JSON Schema subset v1 before an agent trusts or forwards structured output. Returns deterministic machine-readable violations with stable paths; performs no network request, remote $ref resolution, regex evaluation, model inference or mutation. Base mainnet USDC

Deep-compare two caller-supplied JSON values with deterministic object-key ordering and explicit index-based array semantics. Returns bounded added, removed and changed paths with primitive before/after values when safe; performs no network request, model inference, persistence or mutation. Base mainnet USDC

Llama API - pay per call, no API key. OpenAI-format chat completions on meta-llama/llama-4-maverick, pinned to DeepInfra for reliability. You sign a fixed price computed from your max_tokens; unused tokens are not refunded. 20s server-side timeout - never charged if it fires; set your client timeout to 30s. Try GET /llm/llama/sample.

Call OpenAI's gpt-4.1 via a single pay-per-call x402 endpoint — no OpenAI account or API key needed, pay $0.09 per request in USDC. Standard OpenAI chat.completions request/response shape, capped input and output.

Call OpenAI's gpt-5.6-luna via a single pay-per-call x402 endpoint — no OpenAI account or API key needed, pay $0.06 per request in USDC. Standard OpenAI chat.completions request/response shape, capped input and output.

Conform messy or inconsistent JSON to a target schema using Claude Haiku — renames keys, coerces types (string→number, "true"→boolean, etc.), and fills missing fields with null, returning JSON that matches the exact shape the caller specified so agents can normalize data from varied sources into one consistent structure.

Translate many independent strings in one request — each with a caller ID preserved in the response — with per-item status, glossary and tone support, and partial success so one bad item never fails the batch, so agents can localize whole string tables in a single call.

Call OpenAI's gpt-5.4-mini via a single pay-per-call x402 endpoint — no OpenAI account or API key needed, pay $0.04 per request in USDC. Standard OpenAI chat.completions request/response shape, capped input and output.

Convert messy HTML or text into clean, well-structured Markdown using Claude Haiku — strips boilerplate, fixes heading hierarchy, normalizes lists and links, and returns readable Markdown so agents can feed clean docs into downstream pipelines, knowledge bases, or LLM context.

Translate up to 2000 words between languages using Claude Haiku — auto-detects source, supports 30+ target languages, preserves formatting, for larger documents and articles. For short snippets use the cheaper /translate/short.

Repair malformed JSON into valid JSON — fixes code fences, trailing commas, single quotes, unquoted keys, bad escapes, duplicate keys, truncation, and JSON buried in prose. A deterministic tokenizer runs first (no LLM cost), with an optional Haiku fallback, plus optional schema validation and type coercion — so agents can rescue broken model output or dirty data in one call.

Search the Hugging Face Hub for models or datasets (downloads, likes, task, library, tags) or fetch one by id

Fetch historical orderbook snapshots for a Predict.Fun market

Fetch historical orderbook snapshots for a Limitless market

Fetch historical orderbook snapshots for an Opinion market

Purchase a dedicated phone number valid for 30 days. The number is bound to your wallet address. Default country is US (no regulatory friction). Other countries can be requested via the `country` parameter, but Twilio may require a Regulatory Bundle (KYC) on our account before purchase succeeds — if your country fails, contact [email protected] to enable.

mlpeek (Tanod) offline translation between English and 14 languages (Argos Translate, CPU). Use when an agent must translate up to 5,000 characters between English and 14 languages such as Spanish, German or Japanese.

mlpeek (Tanod) speech to text and subtitles: an audio file of up to 10 minutes as text, timed segments, SRT and WebVTT (faster-whisper). Use when an agent must turn a spoken-audio file of up to 10 minutes into text, timed segments or SRT captions.

mlpeek (Tanod) semantic similarity of text pairs (cosine of embeddings). USD 0.0005 per pair, at least USD 0.001 per call (1-50 pairs per call: USD 0.001-0.025). This call: 1 pair.

mlpeek (Tanod) named entities: OntoNotes types with offsets (spaCy, English).

mlpeek (Tanod) text embeddings: 384-dim vectors (bge-small-en or multilingual-e5-small). USD 0.0005 per text, at least USD 0.001 per call (1-64 texts per call: USD 0.001-0.032). This call: 1 text.

mlpeek (Tanod) zero-shot classification into 1-10 caller labels (NLI, English).

mlpeek (Tanod) rerank: up to 100 documents scored against a query (cross-encoder).

Text summarization for agents digesting long documents: condenses audit reports and governance discussions into structured summaries with configurable length and format. Category: inference.

Improve AI search visibility and get cited by ChatGPT, Perplexity, Claude & Google AI Overviews. AEO/GEO/SEO strategy, AI visibility audits, schema markup, llms.txt. 0.05 USDC/query.

Which NAIC insurance model laws a given state has adopted, inverted from the NAIC state-page corpus. Give a state (name or code) and an optional model or topic filter and get the models it has adopted with each classification, citation and linkback.

FDA inspection history for a manufacturer from public FDA records. Give an FDA FEI number or firm name and get the inspection classifications (Official, Voluntary or No Action Indicated) with the trend, plus the Form-483 observation citations with their cited CFR clause and short description. Deterministic, no LLM. Risk indicators, not legal or regulatory advice.

Pick the cheapest LLM that passes your task and constraints. Give a task class and optional constraints such as minimum context, required tool-use, vision or reasoning, maximum input and output price per million tokens, a minimum coding benchmark and a provider filter, and get a ranked cheapest-first list of models that pass, plus a recommended pick and a deterministic rationale. Fused from open keyless catalogs LiteLLM, models dev and Aider with a citation on every model.

OpenAI-compatible chat completion, priced by counted input tokens + max output tokens at per-model rates

Deterministic, paid-per-render SVG bar chart generator for LLM agents. Accepts labeled scalar data and layout options, performs exact geometry math (min/max normalization, nice-interval y-axis ticks, label spacing, bar-width distribution), and returns schema-valid, escape-correct SVG with identical input always yielding byte-identical output.