{
 "schema_version": 1,
 "generated": "2026-10-26",
 "id": "experientiallabs-experiential--turn-agent-traces-into-a-cheaper-router",
 "repo": "experientiallabs/experiential",
 "repo_url": "https://github.com/experientiallabs/experiential",
 "repo_slug": "experientiallabs-experiential",
 "title": "Turn your agent traces into a cheaper router",
 "page_title": "How to turn your agent traces into a cheaper router with Experiential",
 "summary": "Build a simulation from your traces, optimize a router against it and fine-tune a model you own.",
 "who": "An engineer running a support or coding agent in production who wants lower cost without losing quality.",
 "problem": "Picking models by gut feel leaves money on the table, and nobody has data on which cheaper model would have been good enough.",
 "outcome": "A project router fitted to your own traffic, and optionally a fine-tuned open source model.",
 "steps": [
  "Collect OpenTelemetry traces from your current agent, or try the public terminal-tasks OTLP dataset on Hugging Face, which the README downloads with a curl command as traces.otel.jsonl",
  "Run `exp build support-agent` and follow the prompts for providers, models, budget and your trace file",
  "Load the fitted router from Python with `exp.load_router(\"my-project\")` as an official OpenAI client backed by its own private gateway",
  "After collecting traces from your router, run `exp optimize model support-agent` to fine-tune an open source model you own using Tinker"
 ],
 "agent_prompt": "Using experientiallabs/experiential, download the public terminal-tasks trace dataset from the README, run `exp build` on it for a project called <name>, and tell me which models and budget it chose and why. Do not run `exp optimize model` until I approve.",
 "agents": [
  "claude-code",
  "codex"
 ],
 "difficulty": "Advanced",
 "time": "2 hours",
 "basis": "readme",
 "caveat": "The README does not state expected savings, so measure cost and quality on your own traces before you switch traffic.",
 "category": "Software and DevOps",
 "tags": [
  "Software and DevOps"
 ],
 "lang": "python",
 "requirements": [
  "Python with pip to install the package",
  "Provider keys for the models you want to route to (hosted, BYOK or local)",
  "For Capture: macOS, Python 3.13 or newer, and approval for its extension and certificate",
  "Optional: an account on the hosted platform, and Tinker for fine-tuning an open source model"
 ],
 "risk": "none",
 "seen_in_the_wild": false,
 "mentioned_publicly": false,
 "risks": [],
 "best": false,
 "score": 0.8,
 "evidence": [],
 "url": "https://dedupedai.com/use-cases/experientiallabs-experiential/turn-agent-traces-into-a-cheaper-router/",
 "md": "https://dedupedai.com/use-cases/experientiallabs-experiential/turn-agent-traces-into-a-cheaper-router/index.md",
 "disclaimer": "Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked \"Seen in the wild\", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email hello@dedupedai.com and we will look at it promptly.",
 "api": "https://dedupedai.com/api/v1/usecases/experientiallabs-experiential--turn-agent-traces-into-a-cheaper-router.json",
 "api_version": "v1"
}