309,617 packages matching “model-call”
@crewhaus/channel-adapter-base
v0.5.1 · 9 hours ago
Channel boundary base: classifies inbound channel text at TrustOrigin "channel" before it reaches a model call (Pillar 3 chokepoint, Section 33)
No known vulnerabilities
langprotect-armor
v0.1.1 · 28 days ago
Self-hosted LLM observability SDK — trace any model call to your own dashboard
No known vulnerabilities
aqta-verify-receipt
v1.0.8 · 3 days ago
Offline CLI and library for Seal ATTESTATION-v1 receipts. Ed25519 check of a signed model call; no account, no Aqta server required.
No known vulnerabilities
@through-line/adapters-llm
v0.2.179 · 1 month ago
Throughline BYO-LLM helpers: wrap a user-provided model call in a durable step.
No known vulnerabilities
ai-lcr
v0.8.3 · 1 month ago
Least Cost Routing for LLMs — route every model call to the cheapest available provider, fall back automatically, and track real cost. Built for the Vercel AI SDK.
No known vulnerabilities
@crewhaus/compaction-curator
v0.5.1 · 9 hours ago
Pillar-2 active context curation — pre-compaction relevance + dedupe pass that lets compaction-autocompact skip the model call when curation alone brings tokens under the trigger threshold
No known vulnerabilities
@through-line/adapters-ai-sdk
v0.2.179 · 1 month ago
Durable Vercel AI SDK adapter: journal every model call and tool execution as a Throughline step.
No known vulnerabilities
codeglow
v1.0.0 · 13 days ago
Opens a live, interactive architecture design canvas for a repo in your browser, built straight from the source code. No model call, no API key.
No known vulnerabilities
tokenhours-live
v0.3.2 · 13 hours ago
Real-time AI build-cost meter. A localhost proxy that prices every model call live and shows a running total in an instrument HUD — and sees nothing: prompts, responses and keys never leave your machine.
No known vulnerabilities
orion-super-agent-dev
v0.1.12 · 15 hours ago
[dev] Orion — a terminal AI super-agent, strongest at coding. Local brain, local disk: the agent loop runs on your machine; only the metered model call leaves it.
No known vulnerabilities
@voightxyz/openai
v0.1.7 · 2 months ago
Voight observability for the OpenAI SDK. Wrap your OpenAI client and capture every model call — prompts, tokens, costs, latency, errors — surfaced live in the Voight dashboard.
No known vulnerabilities
@loret/sdk
v1.1.1 · 3 months ago
Runtime policy layer for LLM applications — enforce cost, privacy, and runtime guardrails on every model call
No known vulnerabilities
@inkeep/openbolts-canary-echo
v0.0.5 · 3 months ago
Canary plugin used as the upgrade target by @inkeep/openbolts-cli's bun-update live-pair test. Minimal worker-path engine; no model call; no MCP servers.
No known vulnerabilities
@sisu-ai/mw-trace-viewer
v13.0.0 · 4 months ago
Generate interactive HTML and JSON traces to inspect every model call, tool action, and error.
No known vulnerabilities
@thientran01/muse
v0.1.0 · 19 days ago
Point at anything in your running React app, shape it like a design tool, and Muse rewrites the real source. A dev-only visual editing overlay — deterministic, keyless, no model call.
No known vulnerabilities
@davothebigafro/inngest
v0.0.1 · 2 months ago
Optional Inngest add-ons for Shipwright agents: durable tool wrapping, durable model-call types, and OpenTelemetry tracing.
No known vulnerabilities
@pbuda/solidity-mcp
v0.2.0 · 3 months ago
Solidity-native security analysis: MCP server + CLI for Foundry/Hardhat projects. Builds a resolved codebase model, call graph, state-access tracking, mutation maps, invariants, and runs security detectors.
No known vulnerabilities
@ai-sdk/provider
v4.0.6 · 6 hours ago
No description provided.
No known vulnerabilities
@lookspan/openai
v0.4.1 · 2 months ago
Drop-in tracing for the OpenAI SDK — wrap your client in one line and every model call becomes a Lookspan span.
No known vulnerabilities
@ledgerproof/vercel-ai
v1.0.2 · 1 month ago
Vercel AI SDK plug-in that auto-issues EU AI Act Article 50 receipts for every model call.
No known vulnerabilities