{"id":43251149,"url":"https://github.com/verist-ai/verist","last_synced_at":"2026-02-07T18:06:55.478Z","repository":{"id":335725249,"uuid":"1146278163","full_name":"verist-ai/verist","owner":"verist-ai","description":"Replay + diff for AI decisions. Deterministic, audit-first workflow kernel.","archived":false,"fork":false,"pushed_at":"2026-02-04T21:12:23.000Z","size":549,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-02-05T00:58:36.530Z","etag":null,"topics":["ai-infrastructure","ai-systems","ai-workflows","audit-log","compliance","determenistic","developer-tools","diff","explainable-ai","human-in-the-loop","llm","model-evaluation","observability","prompt-engineering","replay","reproducibility","state-machine","traceability","typescript","workflow-engine"],"latest_commit_sha":null,"homepage":"https://verist.dev","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/verist-ai.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":".github/CONTRIBUTING.md","funding":".github/FUNDING.yml","license":"LICENSE","code_of_conduct":".github/CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":".github/SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null},"funding":{"github":"koistya"}},"created_at":"2026-01-30T21:23:33.000Z","updated_at":"2026-02-04T21:11:15.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/verist-ai/verist","commit_stats":null,"previous_names":["verist-ai/verist"],"tags_count":94,"template":false,"template_full_name":null,"purl":"pkg:github/verist-ai/verist","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/verist-ai%2Fverist","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/verist-ai%2Fverist/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/verist-ai%2Fverist/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/verist-ai%2Fverist/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/verist-ai","download_url":"https://codeload.github.com/verist-ai/verist/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/verist-ai%2Fverist/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29203140,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-07T17:44:10.191Z","status":"ssl_error","status_checked_at":"2026-02-07T17:44:07.936Z","response_time":63,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-infrastructure","ai-systems","ai-workflows","audit-log","compliance","determenistic","developer-tools","diff","explainable-ai","human-in-the-loop","llm","model-evaluation","observability","prompt-engineering","replay","reproducibility","state-machine","traceability","typescript","workflow-engine"],"created_at":"2026-02-01T13:07:25.919Z","updated_at":"2026-02-07T18:06:55.453Z","avatar_url":"https://github.com/verist-ai.png","language":"TypeScript","funding_links":["https://github.com/sponsors/koistya"],"categories":[],"sub_categories":[],"readme":"# Verist\n\n[![npm version](https://badge.fury.io/js/@verist%2Fcore.svg)](https://badge.fury.io/js/@verist%2Fcore)\n[![npm downloads](https://img.shields.io/npm/dm/@verist/core.svg)](https://npmjs.com/package/@verist/core)\n[![Ask ChatGPT](https://img.shields.io/badge/Ask_ChatGPT-10a37f?logo=google+gemini\u0026logoColor=white)](https://chatgpt.com/g/g-697e23b923088191b8cb315bebf14a3b-verist-architect)\n[![Twitter Follow](https://img.shields.io/twitter/follow/verist_ai?style=social)](https://x.com/verist_ai)\n\n**Replay + diff for AI decisions — see what changes before you ship.**\n\nA deterministic workflow kernel that captures AI outputs as artifacts, replays them exactly, and shows reviewable diffs when prompts or models change. Human corrections survive recomputation by design.\n\nAsk **Verist Architect** in [ChatGPT](https://chatgpt.com/g/g-697e23b923088191b8cb315bebf14a3b-verist-architect) or [Gemini](https://gemini.google.com/gem/16ofP3wcDulLXY0oj7EXhRlVtws9A8znp?usp=sharing) about replay semantics, diffs, and kernel invariants.\n\n## Why Verist\n\nYou tweak a prompt, upgrade a model, or change extraction logic. AI decisions silently change. A missing field breaks downstream code, a reclassified ticket confuses support, a weaker extraction goes unnoticed — until customers report it.\n\nYou can't answer the question that matters: **\"What decisions would change?\"**\n\n- Logs show what happened, not what _would_ change\n- Testing on 3 examples doesn't catch the 2% regression in production\n- When you recompute, human corrections get overwritten\n\nSo you don't ship. Or you ship and hope.\n\nVerist gives you a different workflow:\n\n```text\nchange prompt → recompute → see diff → approve → ship\n```\n\n- **Diff before shipping** — See what would change after a prompt or model upgrade\n- **Replay any decision** — Reproduce past AI decisions exactly from stored artifacts\n- **Human authority preserved** — Manual overrides survive recomputation by design\n\nVerist is not logging or observability — it is deterministic replay with reviewable diffs.\n\n## Install\n\n```bash\nnpm install @verist/core @verist/replay @verist/cli zod\n```\n\n## Quickstart\n\nGet started with no API keys — `verist init` scaffolds a deterministic step using regex extraction:\n\n```bash\nnpx verist init\nnpx verist capture --step parse-contact --input \"verist/inputs/*.json\"\nnpx verist test --step parse-contact\n```\n\n### Next: LLM Diffs\n\nOnce you have API keys, define an LLM-powered step and see what changes when you modify prompts:\n\n```bash\nOPENAI_API_KEY=sk-... bun examples/prompt-diff/quickstart.ts\n```\n\n```text\n✓ Baseline: 4 claims (specific numbers, names, dates)\n✓ Recompute: 3 claims (vague summaries)\n\n--- Diff ---\n  claims[0]: \"Acme Corp reported $4.2M in Q3 revenue\" → \"Acme Corp had strong Q3 revenue growth\"\n- claims[3]: \"Plans to expand engineering team from 45 to 60 by March 2025\"\n```\n\nFull example: [`examples/prompt-diff/`](./examples/prompt-diff/)\n\n## How It Works\n\nVerist captures AI outputs as artifacts. When you change something, replay against stored inputs and see a diff.\n\n```typescript\nimport { z } from \"zod\";\nimport { defineStep, run, unwrap } from \"@verist/core\";\nimport { recompute, formatDiff } from \"@verist/replay\";\n\n// Define a step with typed input/output\nconst extractClaims = defineStep({\n  name: \"extract-claims\",\n  input: z.object({ text: z.string() }),\n  delta: z.object({ claims: z.array(z.string()) }),\n  run: async (input, ctx) =\u003e {\n    const claims = await ctx.adapters.llm.extract(input.text);\n    return { delta: { claims }, events: [] };\n  },\n});\n\n// Run with artifact capture\nconst result = unwrap(\n  await run(\n    extractClaims,\n    { text },\n    {\n      adapters: { llm },\n      onArtifact: (a) =\u003e store.save(a),\n    },\n  ),\n);\n\n// Later: change prompt, recompute from snapshot\nconst recomputeResult = unwrap(\n  await recompute(snapshot, extractClaims, newCtx),\n);\nconsole.log(formatDiff(recomputeResult.deltaDiff));\n```\n\n## CI Integration\n\nUse `--format json` or `--format markdown` for machine-readable output:\n\n```bash\nverist test --step extract-claims --format json    # structured JSON for scripts\nverist test --step extract-claims --format markdown # PR comment summary\n```\n\nSee [CI Integration Guide](./docs/guides/ci-integration.md) for GitHub Actions examples.\n\n## When to Use\n\n- **Prompt iteration** — Test changes against production history before deploying\n- **Model upgrades** — Quantify impact of switching models (GPT-4 → Claude, etc.)\n- **Safe recompute** — Rerun AI logic without overwriting human corrections\n- **Decision audit** — Replay any past decision exactly, with full provenance\n\nVerist is not a chat framework or agent runtime. It's the trust layer that makes AI decisions reproducible and reviewable.\n\n## Packages\n\n| Package              | Purpose                                       |\n| -------------------- | --------------------------------------------- |\n| `@verist/core`       | Step/workflow definition, execution           |\n| `@verist/replay`     | Artifact capture, replay, recompute with diff |\n| `@verist/cli`        | CLI — replay, diff, and inspect baselines     |\n| `@verist/llm`        | LLM provider adapters with tracing            |\n| `@verist/storage`    | Storage interface and layered state model     |\n| `@verist/storage-pg` | PostgreSQL storage adapter (Drizzle ORM)      |\n| `@verist/artifacts`  | Content-addressable artifact storage          |\n| `@verist/pipeline`   | Sequential step composition                   |\n| `@verist/batch`      | Parallel execution with concurrency control   |\n| `@verist/queue`      | Job queue adapter for distributed execution   |\n| `@verist/otel`       | OpenTelemetry tracing adapter                 |\n\n## Documentation\n\n- [Getting Started](https://verist.dev/getting-started)\n- [CI Integration Guide](./docs/guides/ci-integration.md)\n- [Replay and Diff Guide](https://verist.dev/guides/replay-and-diff)\n\n## Links\n\n[![npm version](https://badge.fury.io/js/@verist%2Fcore.svg)](https://npmjs.com/package/@verist/core)\n[![Twitter](https://img.shields.io/twitter/follow/verist_ai?style=social)](https://x.com/verist_ai)\n\n## License\n\n[Apache-2.0](LICENSE)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fverist-ai%2Fverist","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fverist-ai%2Fverist","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fverist-ai%2Fverist/lists"}