tokenaudit 0.1.0__tar.gz

tokenaudit-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,33 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  checks:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.11", "3.12", "3.13"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+          cache-dependency-path: pyproject.toml
+      - name: Install package and dev tools
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[dev]"
+      - name: Lint
+        run: ruff check .
+      - name: Format check
+        run: ruff format --check .
+      - name: Type check
+        run: mypy
+      - name: Tests
+        run: pytest

tokenaudit-0.1.0/.gitignore

@@ -0,0 +1,24 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+dist/
+build/
+.venv/
+venv/
+
+# Tool caches
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+
+# Environment / editors / OS
+.env
+.vscode/
+.idea/
+.DS_Store
+
+# Capture workdir (throwaway venv + clone, never committed)
+corpus/captured/.work/

tokenaudit-0.1.0/.pre-commit-config.yaml

@@ -0,0 +1,15 @@
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v5.0.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-yaml
+      - id: check-toml
+      - id: check-added-large-files
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.6.9
+    hooks:
+      - id: ruff
+        args: [--fix]
+      - id: ruff-format

tokenaudit-0.1.0/CHANGELOG.md

@@ -0,0 +1,27 @@
+# Changelog
+
+## 0.1.0 — 2026-08-30
+
+First usable release.
+
+- **Ingest** four real trace dialects into one normalized accounting model:
+  OTel GenAI current semconv, legacy flat names (older openllmetry),
+  OpenInference, and the OTel Python SDK's `to_json()` JSONL — with
+  layered-span deduplication (framework + SDK instrumentors double-report
+  the same call), a data-quality gate that refuses to price malformed
+  telemetry, and prompt privacy by construction (block hashes and lengths,
+  never text).
+- **Nine deterministic detectors** — prefix drift, sub-minimum cache
+  blocks, model switch, TTL expiry, breakpoint limit, lookback overflow,
+  tool-schema churn, context bloat, retry storm — each with evidence
+  pointers, bounded impact ranges with stated assumptions, and abstention
+  (with a reason) as a first-class outcome distinct from "checked, clean".
+- **Provider caching rulebook** packaged as sourced, dated JSON (Anthropic,
+  OpenAI, Google), with a CI staleness alarm and a rendered comparison
+  table (`tokenaudit rulebook`).
+- **Evaluation** against a byte-reproducible labeled corpus, published in
+  `eval/results.md`; per-detector corpus records ship inside the package
+  and appear on findings.
+- **CLI**: `tokenaudit diagnose` (text, `--json`, self-contained `--html`
+  report, `--sample`), `tokenaudit ingest`, `tokenaudit rulebook`.
+- Zero runtime dependencies. Python 3.11+.

tokenaudit-0.1.0/DECISIONS.md

@@ -0,0 +1,41 @@
+# Decision log
+
+Short, dated notes on the non-obvious calls, so the "why" survives longer than my memory of it.
+
+## 001 — Verdicts come from deterministic rules, not an LLM (2026-08-29)
+
+**Decision:** the diagnosis engine is a library of explicit, testable rules. No model call anywhere in the verdict path. If an LLM is ever used, it drafts prose or patch text from an already-decided finding, in a module the detectors cannot import.
+
+**Why:** same trace in, same findings out; every detector can be unit-tested against labeled fixtures and given a measured precision; running the tool is free, which matters for a tool whose whole point is reducing cost; and every finding carries an evidence pointer instead of a vibe.
+
+**Cost accepted:** rules only catch what I have enumerated. An enumerated catalog with measured precision beats open-ended guesses I can't test.
+
+## 002 — Name: tokenaudit, not cachelens (2026-08-29)
+
+The working name for this project was "cachelens". The day-one collision check found [cachelens on PyPI](https://pypi.org/project/cachelens/) — released June 2026, an SDK-wrapper cache meter, i.e. a neighbor in this space, not just a name squatter. Shipping a lookalike name would be unfair to them and confusing for everyone.
+
+Picked **tokenaudit** over the other candidates (wastelens, cachedoctor) for two reasons: it leaves the cache-/-lens naming family entirely, and several planned detectors (retry storms, context bloat, tool-schema bloat) are token-waste problems that aren't cache problems at all, so "cache" in the name undersold the scope. Verified free on PyPI and GitHub 2026-08-29.
+
+## 003 — Apache-2.0, not MIT (2026-08-29)
+
+Explicit patent grant, and it matches the ecosystem this project sits in (OpenTelemetry is Apache-2.0).
+
+## 004 — src layout and strict mypy from the first commit (2026-08-29)
+
+src/ layout so tests import the installed package rather than the working directory. `mypy --strict` costs nothing to adopt on an empty package and a lot to adopt later; the detector engine will be pure functions, which is exactly the code strict typing is cheap for.
+
+## 005 — The rulebook is JSON inside the package, not YAML beside it (2026-08-30)
+
+**Decision:** provider caching constants live in `src/tokenaudit/rulebook/*.json`, loaded with the standard library via `importlib.resources`, and ship inside the wheel.
+
+**Why:** the package has zero runtime dependencies, and adding a YAML parser just to read our own constants would end that for no gain — sources and notes are data fields, so YAML's comments aren't needed. Packaging the data means the installed CLI works without a repo checkout. Two boundaries drawn at the same time: trace field-name *mappings* stay in `ingest/dialects.py` (one owner, no code/data drift), and prices appear only where verified — unknown numbers are `null` with a note, never a guess. The corpus generator now reads these same files, so the simulator and the detectors share one set of constants, with the byte-exact corpus regeneration test guarding every edit.
+
+## 006 — argparse over Typer; a ratio over a weighted score; records over probabilities (2026-08-30)
+
+Three calls made together for the 0.1.0 release, all in service of the same principle.
+
+**CLI stays argparse.** Typer gives nicer help text at the cost of the package's zero-runtime-dependency property — a bad trade for a tool whose pitch includes "installs nothing else and never phones home".
+
+**The cache health score is `realized / achievable` savings** — realized cache savings over realized-plus-what-the-findings-quantify. Both halves are computed the same way, so there are no weights to defend; overlapping findings make the denominator conservative, so the score is presented as a floor and says so.
+
+**Findings carry a corpus record, not a confidence number.** A precision float computed from a 16-fixture corpus would be theater. The package ships the raw record instead — "3/4 planted cases found, 0 false alarms in N negatives" — generated by the eval harness and drift-tested, so what appears next to a finding is exactly what was measured.

tokenaudit-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        https://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       https://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

tokenaudit-0.1.0/Makefile

@@ -0,0 +1,10 @@
+.PHONY: corpus check
+
+corpus:
+	python corpus/synthetic/generate.py
+
+check:
+	ruff check .
+	ruff format --check .
+	mypy
+	pytest

tokenaudit-0.1.0/PKG-INFO

@@ -0,0 +1,111 @@
+Metadata-Version: 2.5
+Name: tokenaudit
+Version: 0.1.0
+Summary: Root-cause diagnosis of wasted LLM tokens, from the traces your agents already emit
+Project-URL: Repository, https://github.com/yashmyeole/tokenaudit
+Project-URL: Issues, https://github.com/yashmyeole/tokenaudit/issues
+Author-email: Yash Manohar Yeole <yashmyeole@gmail.com>
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: cost,diagnosis,llm,observability,opentelemetry,prompt-caching,tokens,waste
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: System :: Monitoring
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Provides-Extra: dev
+Requires-Dist: mypy>=1.11; extra == 'dev'
+Requires-Dist: pre-commit>=3.8; extra == 'dev'
+Requires-Dist: pytest>=8.3; extra == 'dev'
+Requires-Dist: ruff>=0.6.9; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# tokenaudit
+
+**Root-cause diagnosis of wasted LLM tokens — from the traces your agents already emit.**
+
+> Status: early development, pre-0.1. Interfaces will change.
+
+Every LLM cost tool can tell you *how much* you spent and *who* spent it. Almost none can tell you *why a specific request wasted money*. The biggest lever on an LLM bill — prompt caching, up to a 10× discount on repeated input — fails silently: change one byte of a cached prefix, drop below a model's minimum cacheable size, let a timer expire, and the provider just bills you full price. No error, no warning. There are [documented](https://barndoor.ai/what-happens-to-your-llm-bill-when-prompt-caching-fails/) [cases](https://github.com/TauricResearch/TradingAgents/issues/750) of agents running at ~0% cache hit rate for months, at 30–40% unnecessary cost, because of a single line of prompt construction.
+
+tokenaudit reads the traces agents already produce and, for each expensive request, answers four questions: **what was wasted, why, what it cost, and what to change.**
+
+## Quickstart
+
+    pip install tokenaudit
+    tokenaudit diagnose --sample          # diagnose the bundled demo trace
+    tokenaudit diagnose mytraces/ --html report.html
+
+Zero runtime dependencies; reads OTLP/JSON files and OTel SDK JSONL. Try it
+on this repo's real captured runs: `tokenaudit diagnose corpus/captured/runs`.
+
+The plan, in one line each:
+
+- Ingest standard traces — OpenTelemetry GenAI in all three of its cache-field naming generations, OpenInference, raw provider SDK logs — into one normalized accounting model.
+- Run a library of deterministic detectors for the silent failure modes: prefix drift, sub-minimum cache blocks, model switches, TTL expiry, breakpoint and lookback limits, tool-schema churn and bloat, context bloat, retry storms.
+- Price each finding against a counterfactual — what the bill would have been under the fix — as a bounded range, never a point guess.
+- Publish per-detector precision and recall against a labeled corpus, so you know how much to trust each finding.
+
+## Prior art
+
+This space is crowded, and being precise about what already exists is the point of this section.
+
+| Tool | What it does |
+|---|---|
+| [CacheLens](https://github.com/ChingEnLin/CacheLens) | Closest prior art. Dev-time SDK wrapper that meters cache hit rate and prefix stability per prompt layer in apps you instrument. Doesn't read existing traces, and prefix stability is its one diagnosis. |
+| [Helicone](https://helicone.ai) | Observability proxy: cost and token accounting per user/session/feature, including cache token counts |
+| [Langfuse](https://langfuse.com) | Open-source LLM engineering platform: tracing, evals, per-trace cost accounting |
+| [Datadog LLM Observability](https://docs.datadoghq.com/llm_observability/) | Enterprise monitoring: per-request token and cost tracking, quality-issue clustering |
+| [Braintrust](https://braintrust.dev) | Evals and observability; its Loop assistant root-causes *quality* failures |
+| [LiteLLM](https://github.com/BerriAI/litellm) | Gateway with budgets, rate limits and spend tracking across providers |
+| [ccusage](https://github.com/ryoppippi/ccusage) | CLI usage and cost reporting from Claude Code's local logs |
+| [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/) | Platform-level caching, spend caps and alerts |
+
+All of these answer "how much", and several now count cache tokens. The ones that do automatic root-cause analysis target response *quality*, not cost. The closest tool diagnoses one failure mode, for one integration style, going forward only. What none of them do is take the telemetry any agent already emitted — no code changes, any vendor — and say: *these 44 requests missed cache because the system prompt embeds a timestamp; here is the evidence, and here is what it cost.* That attribution step is this project.
+
+If you know a tool that already does this, please open an issue — I would genuinely rather find out now.
+
+## What tokenaudit is not
+
+- **Not a dashboard.** The tools above already report spend well.
+- **Not a proxy, gateway, or wrapper.** It reads traces after the fact and never sits in the request path, so it can't change what it measures and doesn't need your code.
+- **Not an LLM reading your logs.** Verdicts come from explicit rules with unit tests and a published false-positive rate ([decision 001](DECISIONS.md)).
+- **Not a savings guarantee.** Estimates are ranges with stated assumptions, and requests whose telemetry fails a plausibility check are flagged as unpriceable, not silently priced.
+
+## Planned layout
+
+    src/tokenaudit/
+      ingest/          trace readers + schema normalization
+      model/           normalized requests/sessions + token accounting
+      detect/          the diagnosis engine (deterministic detectors)
+      counterfactual/  what the bill would have been under a fix
+      report/          CLI, JSON and HTML output
+      ci/              cost-regression gate for pipelines
+
+## Roadmap
+
+- [x] Prior-art survey, positioning, project scaffold
+- [x] Labeled trace corpus: synthetic known-answer fixtures + captured runs from real public agents
+- [x] Trace reader and normalizer with a data-quality gate (bad telemetry is refused, not zero-filled)
+- [x] Provider caching rulebook: minimums, TTL semantics, invalidation rules — versioned, with sources
+- [x] Detectors, starting with prefix drift, sub-minimum blocks, and model/version switch
+- [x] Evaluation harness: per-detector results published in [eval/results.md](eval/results.md)
+- [x] `tokenaudit diagnose` CLI, HTML report, cache health score
+- [ ] CI cost-regression mode: fail a build that tanks the cache hit rate
+- [ ] Case studies on public agents, with reproduction commands
+
+## Development
+
+    python -m venv .venv && source .venv/bin/activate
+    pip install -e ".[dev]"
+    pre-commit install
+    pytest
+
+## License
+
+Apache-2.0

tokenaudit-0.1.0/README.md

@@ -0,0 +1,84 @@
+# tokenaudit
+
+**Root-cause diagnosis of wasted LLM tokens — from the traces your agents already emit.**
+
+> Status: early development, pre-0.1. Interfaces will change.
+
+Every LLM cost tool can tell you *how much* you spent and *who* spent it. Almost none can tell you *why a specific request wasted money*. The biggest lever on an LLM bill — prompt caching, up to a 10× discount on repeated input — fails silently: change one byte of a cached prefix, drop below a model's minimum cacheable size, let a timer expire, and the provider just bills you full price. No error, no warning. There are [documented](https://barndoor.ai/what-happens-to-your-llm-bill-when-prompt-caching-fails/) [cases](https://github.com/TauricResearch/TradingAgents/issues/750) of agents running at ~0% cache hit rate for months, at 30–40% unnecessary cost, because of a single line of prompt construction.
+
+tokenaudit reads the traces agents already produce and, for each expensive request, answers four questions: **what was wasted, why, what it cost, and what to change.**
+
+## Quickstart
+
+    pip install tokenaudit
+    tokenaudit diagnose --sample          # diagnose the bundled demo trace
+    tokenaudit diagnose mytraces/ --html report.html
+
+Zero runtime dependencies; reads OTLP/JSON files and OTel SDK JSONL. Try it
+on this repo's real captured runs: `tokenaudit diagnose corpus/captured/runs`.
+
+The plan, in one line each:
+
+- Ingest standard traces — OpenTelemetry GenAI in all three of its cache-field naming generations, OpenInference, raw provider SDK logs — into one normalized accounting model.
+- Run a library of deterministic detectors for the silent failure modes: prefix drift, sub-minimum cache blocks, model switches, TTL expiry, breakpoint and lookback limits, tool-schema churn and bloat, context bloat, retry storms.
+- Price each finding against a counterfactual — what the bill would have been under the fix — as a bounded range, never a point guess.
+- Publish per-detector precision and recall against a labeled corpus, so you know how much to trust each finding.
+
+## Prior art
+
+This space is crowded, and being precise about what already exists is the point of this section.
+
+| Tool | What it does |
+|---|---|
+| [CacheLens](https://github.com/ChingEnLin/CacheLens) | Closest prior art. Dev-time SDK wrapper that meters cache hit rate and prefix stability per prompt layer in apps you instrument. Doesn't read existing traces, and prefix stability is its one diagnosis. |
+| [Helicone](https://helicone.ai) | Observability proxy: cost and token accounting per user/session/feature, including cache token counts |
+| [Langfuse](https://langfuse.com) | Open-source LLM engineering platform: tracing, evals, per-trace cost accounting |
+| [Datadog LLM Observability](https://docs.datadoghq.com/llm_observability/) | Enterprise monitoring: per-request token and cost tracking, quality-issue clustering |
+| [Braintrust](https://braintrust.dev) | Evals and observability; its Loop assistant root-causes *quality* failures |
+| [LiteLLM](https://github.com/BerriAI/litellm) | Gateway with budgets, rate limits and spend tracking across providers |
+| [ccusage](https://github.com/ryoppippi/ccusage) | CLI usage and cost reporting from Claude Code's local logs |
+| [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/) | Platform-level caching, spend caps and alerts |
+
+All of these answer "how much", and several now count cache tokens. The ones that do automatic root-cause analysis target response *quality*, not cost. The closest tool diagnoses one failure mode, for one integration style, going forward only. What none of them do is take the telemetry any agent already emitted — no code changes, any vendor — and say: *these 44 requests missed cache because the system prompt embeds a timestamp; here is the evidence, and here is what it cost.* That attribution step is this project.
+
+If you know a tool that already does this, please open an issue — I would genuinely rather find out now.
+
+## What tokenaudit is not
+
+- **Not a dashboard.** The tools above already report spend well.
+- **Not a proxy, gateway, or wrapper.** It reads traces after the fact and never sits in the request path, so it can't change what it measures and doesn't need your code.
+- **Not an LLM reading your logs.** Verdicts come from explicit rules with unit tests and a published false-positive rate ([decision 001](DECISIONS.md)).
+- **Not a savings guarantee.** Estimates are ranges with stated assumptions, and requests whose telemetry fails a plausibility check are flagged as unpriceable, not silently priced.
+
+## Planned layout
+
+    src/tokenaudit/
+      ingest/          trace readers + schema normalization
+      model/           normalized requests/sessions + token accounting
+      detect/          the diagnosis engine (deterministic detectors)
+      counterfactual/  what the bill would have been under a fix
+      report/          CLI, JSON and HTML output
+      ci/              cost-regression gate for pipelines
+
+## Roadmap
+
+- [x] Prior-art survey, positioning, project scaffold
+- [x] Labeled trace corpus: synthetic known-answer fixtures + captured runs from real public agents
+- [x] Trace reader and normalizer with a data-quality gate (bad telemetry is refused, not zero-filled)
+- [x] Provider caching rulebook: minimums, TTL semantics, invalidation rules — versioned, with sources
+- [x] Detectors, starting with prefix drift, sub-minimum blocks, and model/version switch
+- [x] Evaluation harness: per-detector results published in [eval/results.md](eval/results.md)
+- [x] `tokenaudit diagnose` CLI, HTML report, cache health score
+- [ ] CI cost-regression mode: fail a build that tanks the cache hit rate
+- [ ] Case studies on public agents, with reproduction commands
+
+## Development
+
+    python -m venv .venv && source .venv/bin/activate
+    pip install -e ".[dev]"
+    pre-commit install
+    pytest
+
+## License
+
+Apache-2.0

tokenaudit-0.1.0/corpus/README.md

@@ -0,0 +1,83 @@
+# The corpus
+
+Ground truth for the diagnosis engine, in two tracks:
+
+- `synthetic/` — generated fixtures with **known answers**. A deterministic
+  simulator (`synthetic/generate.py`) replays agent sessions under specific
+  broken configurations and emits the traces a correctly instrumented app
+  would have produced, in three real schema dialects.
+- `captured/` — recorded runs of **real public projects** (arrives with the
+  capture track). These contain genuine waste nobody scripted.
+
+## The boundary between the tracks
+
+Synthetic fixtures are **unit tests, never evidence**. "The tool catches
+nine bugs I planted myself" is a strawman, and this project doesn't ship
+strawmen. Any number that appears in the top-level README — savings, hit
+rates, case-study results — must come from `captured/` runs of code I did
+not write. The synthetic track exists so that every detector has a measured
+precision and recall against cases with published answers, including
+negative cases where a correct tool must stay silent.
+
+One caveat worth stating plainly: the simulator encodes the same provider
+rules the detectors will check, so agreement between them is partly
+circular. That is what makes the captured track non-optional — real traces
+are the check against reality, and the simulator's rules cite the provider
+docs they came from so they can be audited independently.
+
+## Label format
+
+Every `<fixture>.trace.json` ships with `<fixture>.label.json`:
+
+| Field | Meaning |
+|---|---|
+| `contains` | failure modes genuinely present — the tool should report these |
+| `must_not_contain` | every other mode — reporting one of these is a false positive |
+| `content_captured` | whether message content is in the trace; without it, a tool can prove waste but must not guess the cause |
+| `evidence` | machine-readable pointers (which call, which block, which gap) |
+| `totals` | session-level token accounting, used to cross-check dialects |
+| `notes` | the human explanation |
+
+`contains` and `must_not_contain` always partition the full failure-mode
+set, so an evaluation can score both misses and false alarms.
+
+**Attribution precedence:** when tool definitions change or reorder, ground
+truth is `tool_schema_churn`, not `prefix_drift`, even though the prefix
+did technically change — the fix is different, so the finding must be too.
+Fixtures encode this: the tool-churn fixture lists `prefix_drift` under
+`must_not_contain`.
+
+## Determinism
+
+The generator uses a fixed simulated clock (2026-01-15T09:00:00Z), no
+randomness, integer-only payloads, and stable JSON rendering. The same
+command produces byte-identical output on any machine and Python ≥ 3.10:
+
+    python corpus/synthetic/generate.py
+
+`tests/test_corpus.py` regenerates the corpus on every test run and fails
+if the committed files differ from what the generator produces, so the
+generator and fixtures cannot drift apart. Every trace carries
+`tokenaudit.synthetic = true` in its resource attributes so a synthetic
+file can never be mistaken for a captured one.
+
+## Schema dialects covered
+
+1. **Current OTel GenAI semconv** — `gen_ai.usage.cache_read.input_tokens`,
+   `gen_ai.usage.cache_creation.input_tokens`; `gen_ai.usage.input_tokens`
+   includes cached tokens. Message content (opt-in in real
+   instrumentations) rides `gen_ai.input.messages` / `gen_ai.output.messages`.
+2. **Legacy flat names** (older openllmetry) —
+   `gen_ai.usage.cache_read_input_tokens` etc., alongside deprecated
+   `gen_ai.usage.prompt_tokens`, which passes the provider's field through —
+   for Anthropic that count EXCLUDES cache tokens. Same session, different
+   arithmetic: the normalizer has to make them agree.
+3. **OpenInference** — `llm.token_count.prompt_details.cache_read` /
+   `.cache_write`, flattened `llm.input_messages.<i>.message.*` attributes.
+
+Sources (verified 2026-08-29):
+[OTel GenAI spans](https://github.com/open-telemetry/semantic-conventions-genai/blob/main/docs/gen-ai/gen-ai-spans.md) ·
+[openllmetry #3647](https://github.com/traceloop/openllmetry/issues/3647) ·
+[OpenInference spec](https://github.com/Arize-ai/openinference/blob/main/spec/semantic_conventions.md) ·
+[Anthropic prompt caching](https://platform.claude.com/docs/en/build-with-claude/prompt-caching) ·
+[OpenAI prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching)