All projects matching “llm”

LLM chat completion with no API key or signup (fast open model, meta-llama/llama-3.3-70b-instruct). OpenAI-compatible: set your SDK base URL to https://keyless-theta.vercel.app/fast/v1. Flat $0.002 per call, up to 1500 output tokens. Charged only on success.

Cheapest LLM meeting your constraints (price/context/modality) PLUS every price change and deprecation in the window. Providers publish only today's catalog and overwrite it — the change history exists nowhere else

Deepent Research — refusal-free LLM inference (per-request)

Fetches a public web page URL and returns its main article content as clean Markdown with title, byline and excerpt; strips navigation, ads and scripts. Useful for RAG ingestion and LLM context.

Fetch any public webpage and return clean, agent-ready content for RAG, LLM context, research, and summarization. Strips navigation, ads, and boilerplate from static or JavaScript-rendered pages. Supports markdown, plain text, or structured JSON output.

Verify up to 50 citations in one call, one flat price. Body: {"citations": ["<string>", ...]} - same shapes and verdicts as the single route. Returns results[] (one per citation), the verified/not_found/suspicious counts, verification_time_ms, hallucination_risk (LOW all verified, MEDIUM 1-2 not found, HIGH 3+). Deterministic, no LLM in the path. suspicious = a name_match mismatch: a DIFFERENT cas

Extrae el contenido de una URL y lo devuelve limpio en Markdown/HTML para el contexto de un LLM.

Agent Reputation Check by Agentropolis: give a payTo and/or an agent / A2A / MCP URL. Returns owner, verified domain, real outside buyer history, wash flags and trust score from Agentropolis. Says when coverage is limited. Deterministic, no LLM.

Renders JavaScript-heavy SPA pages in real headless Chromium via Cloudflare Browser Rendering, waits for network-idle / selector-ready, strips nav/footer/script chrome, and serializes the fully-rendered DOM into clean, LLM-ready Markdown.

Claim check with citations (claims only, never people). Wikidata+OpenAlex+Crossref grounding + LLM verdicts. Unverifiable → charged:false.

AI Web Extractor — turn ANY url into clean, LLM-ready markdown. POST {url}; get the main content as markdown (boilerplate/nav/ads stripped) with title, description, and links. The scrape / read / url-to-markdown primitive for RAG and research agents. Fast, cheap, x402-native.

Count the LLM tokens in text for a given model or encoding — exact for OpenAI and open models (Llama, Mistral, Qwen, Gemma, DeepSeek, Command-R); labelled-approximate for Claude & Gemini.

Fuzzy-dedupe a list of records into clusters of likely-duplicate entities. Blocks by normalized token prefix, scores with Jaro-Winkler + token-set matching (exact on email/phone), unions matches above threshold, and returns a merged canonical record per cluster with a confidence score. Deterministic, no LLM calls.

Detect and redact PII in text or JSON. Recognizers are pure pattern/checksum matching (regex + Luhn/mod-97/SSN-area validation) — no LLM calls, no external services. Person names and dates of birth are matched only via nearby context (e.g. 'Mr. Smith', 'DOB: ...'), not general NER, to keep false positives low.

Rug-risk verdict for a Base ERC20 address: 0-100 score, A-F grade, one-line summary, graded checks (ERC20, honeypot selectors, liquidity, market) plus a separate ungraded adminPowers profile (proxy, owner, mint, blacklist, pause), each with raw evidence. No LLM.

Onchain facts for a Base ERC20 address: ERC20 shape, honeypot-style selectors and an ungraded adminPowers profile (proxy, owner, mint, blacklist, pause) with raw evidence per check. No DEX data, grade is n/a. No LLM.

OpenAI-compatible LLM Chat Completions Gateway ($0.002 USDC/call). Drop-in replacement for /v1/chat/completions — change base_url to https://agent-kiosk.vercel.app/v1 and pay per request in Base USDC via x402 with zero account or API key setup.

Should your agent buy this Base token? Half the price of $0.02 AI token-verdict quests, deterministic on-chain checks instead of an LLM opinion. One-call verdict OK|CAUTION|AVOID + should_buy + can_sell + plain-English reasons, backed by risk_score 0-100 + level, flags (honeypot, cannot_sell_all, high tax, mintable, hidden owner, proxy, blacklist, unverified), bytecode size, EIP-1967 proxy impl, o

Agent data APIs: web search, page crawl, company enrich, X/Twitter mentions — plus cheap LLM. GET any URL → HTTP 402 USDC Solana. Pay, retry PAYMENT-SIGNATURE, get JSON. search, crawl, enrich, twitter, inference. Models: openrouter/free $0.001, google/gemma-4-31b-it:free $0.001, google/gemma-4-26b-a4b-it:free $0.001, n

Translate text between languages (LLM, structure and placeholders preserved) (1 × 1,000 characters at 0.030000 USDC each)

Compact LLM-friendly Hyperliquid perp market brief in one call: a plain-English brief_text paragraph plus JSON for BTC/ETH/SOL/HYPE (mark, 24h change, funding APR, predicted funding, OI and OI change), top-5 funding extremes each side, largest whale net tilts, nearest whale liquidation levels for the majors, and HL listing/delisting events with our first-seen timestamps. Freshness fields included.

Fetch portfolio balances and LLM-generated strategies

Fetch any public web page and return clean, LLM-ready Markdown of the main content (boilerplate, nav and ads removed) plus title, byline, site name, language, publish date and word count. Optional absolute link list.

Indonesian market sentiment index for one topic: a 0-100 score (50 = neutral), label, risk-on/risk-off mood, counts per sentiment label, top positive and negative driver terms, entities most associated with each direction, and the strongest headlines on both sides. Deterministic lexicon, no LLM — every article in /headlines carries sentiment.matches, so the number can be audited. Useful for tradin

Web page to Markdown, $0.001 USDC per call: Fetch a public web page and return LLM-ready Markdown: title, meta description, canonical, language, main content (scripts/nav/footer stripped, headings/lists/links kept), first 50 absolute links, word count. Honours robots.txt (disallowed → 403, not charged). No JS rendering; non-HTML → 415. max_chars 500-50000 (default 20000). No params = runs the exam

Converts one LLM tool/function definition (OpenAI, Anthropic, Gemini, MCP or plain JSON Schema; format auto-detected) into one or more target provider formats. Applies each provider's schema rules: OpenAI strict mode (all fields required, optional fields made nullable, additionalProperties false), Gemini's OpenAPI subset (no $ref, nullable, string enums), $ref inlining and name sanitising. Returns

Deterministic A2A Broken-JSON Repair & Schema Validator (/v1/json-repair) — Repairs truncated or malformed LLM JSON outputs (trailing commas, markdown fences, single quotes) so multi-agent pipelines never crash. (https://json-repair.agentdiet.dev/v1/json-repair)

Schema-Healer Deterministic Malformed JSON Repair Node: Repairs truncated or malformed LLM JSON outputs into strict schema-compliant objects.

Extract clean, LLM-ready Markdown or token-efficient text from any public webpage for RAG ingestion, research, summarization, or agent context. Strips navigation, ads, scripts, forms, and repeated boilerplate; preserves headings and resolved links; labels fetched content as untrusted; and returns at most 50,000 characters.

Fetch a public web page and return clean LLM-ready Markdown (Readability main-content extraction, nav/scripts stripped) plus title, byline, excerpt and OpenGraph metadata. {fullPage:true} keeps the whole body; {maxChars} caps output (default 100k). No JS rendering.