buddie 1.0.0__tar.gz

buddie-1.0.0/CHANGELOG.md

@@ -0,0 +1,232 @@
+# Changelog
+
+## buddie 1.0.0 (2026-10-19)
+
+First stable release: the receipt, the CLI, the MCP tools and `buddie.toml` follow semantic versioning from here on.
+
+New places to run:
+- Claude Desktop: an MCP Bundle (`manifest.json`, `.mcpb` built and attached to each release),
+  `docs/install-claude-desktop.md`.
+- Codex: plugin manifest `.codex-plugin/plugin.json` with the MCP server, the skills and a Stop hook; this repository
+  is a Codex marketplace too. `docs/install-codex.md`.
+- ChatGPT, claude.ai and the Claude API: a remote MCP server over Streamable HTTP (`buddie serve --http`) with a
+  narrower tool set and a fetch guard (public addresses only, every redirect checked), a Docker image,
+  `docs/deploy.md`.
+- PyPI: `pip install buddie`, `uvx buddie serve`; release workflow with trusted publishing, MCP Registry entry
+  `io.github.tsyyan/buddie` (`server.json`).
+
+Changes:
+- **Breaking:** the quote module is `buddie.verbatim` and its command `buddie-verbatim`. The top-level `verbatim`
+  package and command clashed with an unrelated `verbatim` on PyPI.
+- The final answer at `Stop` comes from the event's `last_assistant_message` when the transcript has none (Codex).
+- `run:` against a transcript that is not Claude Code's is `UNCHECKABLE`, not a false `BROKEN`.
+- `BUDDIE_DEFAULT_REPOS=none` stops `verify_report` from defaulting to the repo around the server's cwd.
+- The Claude Code plugin asks for an optional GitHub token (`userConfig`, stored as a secret) instead of relying on
+  `GITHUB_TOKEN` in the environment; an unset option is never sent.
+- MCP tool annotations (read-only, non-destructive) for clients and directories.
+- Development moved here from the private lab repository: this repository is the source of buddie, no longer an
+  export. `PROVENANCE.json` moved to `results/` as the record of the last export (`lab@ed8d971`).
+- CI: `apt-get update` before installing poppler-utils; tests import their sibling modules directly, so bare
+  `pytest` collects them; jobs for the package, the bundle, the Docker image and the catalog manifests.
+
+- Development moved here from the private lab repository: this repository is the source of buddie, no longer an
+  export. `PROVENANCE.json` moved to `results/` as the record of the last export (`lab@ed8d971`).
+- CI: `apt-get update` before installing poppler-utils; tests import their sibling modules directly, so bare
+  `pytest` collects them.
+
+## buddie 0.8.2 (2026-10-08)
+
+- Fewer false returns for work claimed in prose. A word in a clause opened by once, as soon as, after, when, if,
+  until, unless or before (Russian «когда», «если», «как только», «после того как», «пока не», «до того как»), within
+  three words and no punctuation, is a condition or a plan, not a claim: «Once automerge has merged it, start the
+  next row». «merged» after an article or possessive and before a noun is an adjective: «the merged code».
+- At Stop, the last message is not checked for prose when the same turn already sent a final answer through a text
+  tool (reply, post_message, send_message) that the gate let out: that text reaches no one. A reply the gate
+  returned, or one sent in an earlier turn, does not count.
+- In lab's journal since prose started blocking, these were 2 of 8 prose returns read from transcripts (false) and
+  3 of 8 on Stop text after the reply had gone out.
+
+## buddie 0.8.1 (2026-10-04)
+
+- A bare anchor counts: after the anchor sign, without backticks, `path#sha`, `path@commit`, `fact=value`, `pr:…`,
+  `ci:…`, `choice:…` up to the first space, `run:CMD => OUT` up to the end of the line, the next sign or a table cell
+  border; before, only an anchor in backticks was found, and a report anchored the way the SKILL.md table shows was
+  returned as "final answer has no anchor". Placeholders (`<…>`, `…`, `OWNER/REPO`, `CMD => OUT`) and a sign with no
+  anchor after it are not anchors. On the 43 E017 gate rounds, offline: rounds with the sign returned for no anchor
+  14 → 0.
+- The return no longer offers a ready anchor from a call that has nothing to do with the claim: a command that opens by
+  writing a file (a heredoc into `>` or `tee`, `sed -i`, `perl -pi`) or the `SubagentHandback` call itself.
+
+## buddie 0.8.0 (2026-10-04)
+
+- `buddie.toml` `[gates]`: each class of claims (`anchors`, `quotes`, `final`, `prose`, `numbers`, `work`, `summary`,
+  `promises`) is `block`, `warn` or `off`. A warning lets the text out with a `systemMessage` naming what the gate
+  found. `BUDDIE_GATES` overrides the file. Defaults change: work claimed in prose and numbers missing from their
+  source warn instead of blocking or being only a gap; the subagent summary and the promise check are off.
+- `buddie init` writes `buddie.toml` with the defaults and the intro mode: until 20 final answers have passed with no
+  false return reported (`buddie intro --false`), nothing is returned, each would-be return is a warning.
+  `buddie enforce` ends it.
+- Windows: the hooks start through `hooks/run.sh` (`python3`, else `python`, else `py -3`); the hook reads its event
+  as UTF-8 and writes UTF-8. CI runs the tests and a smoke run of the hook on recorded events on Linux, macOS and
+  Windows, from a clone and from `pip install`.
+
+## buddie 0.7.4 (2026-10-04)
+
+- Defaults for other projects. Without a configured mark the final answer is the last assistant message at `Stop`,
+  in any language; before, only text carrying the Russian mark `Следующий шаг:` was final, so the final-answer rules
+  (an anchor required, no count or work claim in prose without one) never ran for anyone else. A project that hands
+  in results through a tool names its mark as `final_mark` in `buddie.toml` `[mandate]` or in `BUDDIE_FINAL`; then
+  text carrying it is final at any event. An empty `BUDDIE_FINAL` turns the rules off.
+- The plugin's hook matches Claude Code's own tools only (`SendMessage`; `Agent` and `Task` for launches).
+  Matchers for a project's reply and post tools and for cloud sessions moved to
+  [`examples/hooks/projects.json`](examples/hooks/projects.json).
+- The version in `pyproject.toml` and `CITATION.cff` is written from the code at export: the public tree carried
+  0.6.0 under 0.7.3 code, and `test_one_version_everywhere` failed.
+
+## buddie 0.7.3 (2026-10-04)
+
+- The launch gate and the summary judge a row's verdict at the time of the check: a "not before <date>" condition
+  comes due then, and a delay ("a day after the merge of №M") counts from the commit that brought the row's text.
+  Without history (uncommitted edit, the bottom of a shallow clone) a delay does not come due.
+
+## buddie 0.7.2 (2026-10-04)
+
+- A launch with a basis is still returned (exit 2, outcome `taken`) when the row is already taken: `free: false` in
+  the row's verdict of the merge comment on the PR merged as the anchor's commit, or the command `[launch] free`
+  (`{n}` is the row) answering `free: false` now. Its reasons are printed. `BUDDIE_FREE=<command>` sets the command,
+  `0` turns it off; without a comment and a command the gate lets the launch through (`free: null` in the journal).
+- The launch hook's timeout is 60 s instead of 30 s.
+
+## buddie 0.7.0 (2026-10-04)
+
+- Launch gate (`buddie.toml` `[launch] tools`, off without the section): a launch of queue row N passes only with
+  `⚓ NEXT.md@<commit>` whose queue gives N the verdict `auto`, or with `⚓ choice:<id>=№N`, the person's pick in the
+  journal; otherwise exit 2 with a ready anchor when one holds. A launch with no row number passes.
+- Promises: a text the gate let out is read for promises to launch a row (with a condition: a PR merged, a row done)
+  and to send a summary; `SessionStart` prints the open ones, `Stop` returns the turn once per promise whose
+  condition holds. A `buddie:` line drops a promise; promises expire after 48 hours.
+
+## verbatim 0.3.7 (2026-10-04)
+
+- `numbers` reads "млн", "млрд", "тыс.", "трлн" and the full Russian words after a number as million, billion,
+  thousand, trillion: "$35.0 млн" is found on a page with "$35.0 million".
+- A year the sentence itself calls absent ("not on the page", "не упоминается") is not looked up (`ABSENT`).
+- A report on one's own work is not a claim about the page: numbers inside a `run:` anchor (command and output) and a
+  snapshot timestamp next to its sha256 or a word like "snapshot" are not looked up.
+
+## verbatim 0.3.6 (2026-10-04)
+
+- `numbers`: a number missing from an abstract page (arXiv `/abs/`, an OpenAlex abstract) is `UNCERTAIN`
+  (`abstract_only`); one missing from a market report whose forecast period on the page is later than the sentence's
+  is `UNCERTAIN` (`rewritten`). buddie counts both as gaps, not blocks.
+
+## verbatim 0.3.5 (2026-10-03)
+
+- `check.readable` no longer takes for a document an Anubis check, a login page, an encoded payload or a redirect to a
+  local address.
+- `numbers` reads "2024 Jan 5", "10/09/2018" (both orders), thin-space thousands and tables "in thousands"/"in
+  millions"; "from 12.3 million in 2020 to 16.2 million" is no longer a range.
+
+## verbatim 0.3.4 (2026-10-03)
+
+- `check.readable` no longer takes for a document a "Client Challenge", a captcha at the head of a long page, a
+  JavaScript shell, or a redirect to the home page, to a section above the requested page or to a short page of
+  another site; the `--access` cascade then goes on to the real page.
+- `numbers`: a year counts as found only next to a word of its own clause; `cited_sentences` reads Gemini-style
+  footnotes.
+
+## buddie 0.7.1 (2026-10-04)
+
+- A final answer that claims work in prose (merged, CI green, tests passed, ran) with no `pr:`, `ci:` or `run:`
+  anchor on the same line is returned, with the ready anchors that hold now; a `buddie:` line does not let it
+  through, only an anchor or dropping the claim does. Other messages only count such lines (`prose_work`); the
+  receipt now also lists their numbers (`prose_lines`). A negated claim ("not merged yet", «ещё не слит») is not a
+  claim. Found in our own dogfood: a thread answer stated four merges with no anchor and left.
+
+## buddie 0.6.1 (2026-10-03)
+
+- A line whose quote is FOUND on the page at a URL written on that same line carries its source: its numbers are no
+  longer listed under `unanchored`, and a final answer is not returned for them as "number without anchor". The
+  receipt lists such lines under `quote_backed`. A number without the URL on its line is still returned. Found on a
+  correct report of the E016 pilot (three quoted revenue figures, quotes 3/3 found, returned anyway).
+
+## buddie 0.6.0 (2026-10-03)
+
+- Summary with a choice card: the hook counts live subagents (`SubagentStart`/`SubagentStop`), checks each
+  subagent's last message, and on the orchestrator's `Stop` with none left asks once for a summary.
+- `buddie summary` (CLI) and the MCP tool `summary`: what was done (receipt outcome), lost subagents, next steps with
+  `auto`/`ask` verdicts, open PRs and red CI, and a ready `AskUserQuestion` card with at most three steps.
+- The person's pick (`PostToolUse` on `AskUserQuestion`) goes to the journal; the anchor `choice:<id>=<step>` is
+  checked against it and counts as consent in work claims.
+
+## buddie 0.5.1 (2026-10-03)
+
+- An anchor with the sign before the backticks (⚓ `path@commit`, as the skill used to show it) is now read like
+  `⚓ path@commit`. Before, it was not parsed at all: a correct report counted as unanchored and was returned, and a
+  broken `pr:`, `ci:` or `run:` anchor in that form went through unchecked. The skill and README show the sign inside
+  the backticks.
+
+## buddie 0.5.0 (2026-10-03)
+
+The public tree had carried 0.4 and 0.5 code under the 0.3.0 version since the daily sync; this release names it.
+
+- Numbers (0.4): numbers and dates in a sentence with a link are looked up in the snapshot of that link
+  (`verbatim.numbers`). A number missing from its readable source is a gap by default; `--numbers fail` or
+  `BUDDIE_NUMBERS=fail` makes it block, `off` turns it off.
+- Closed loop (0.5): a returned message logs `reasons`; the next attempt logs `gate`, how it got through (proof,
+  removal, disclosure, returned again). `SessionStart` prints the lessons of past sessions' journal (at most ten lines,
+  `BUDDIE_LESSONS=0` turns it off). A return offers ready `pr:`, `ci:` and `run:` anchors that already hold.
+  `buddie journal measure` and `buddie journal lessons`.
+- Context gate: past `BUDDIE_CONTEXT_LIMIT` tokens (default 200000) a non-final reply is returned with a reminder to
+  hand the result over and start the next step in a new thread.
+- Snapshot stores: the hook judges web quotes against the snapshots the session took (`BUDDIE_STORES`,
+  `VERBATIM_STORE`, stores named in the session's tool calls, `./.verbatim`), and fetches only URLs in none of them.
+- A `buddie:` disclosure line no longer lets a broken `pr:`, `ci:` or `run:` anchor through: such an anchor states work
+  that did not happen, so it must be corrected or removed.
+
+## buddie 0.3.0 (2026-10-03), first public release
+
+The repository `tsyyan/verbatim` became `tsyyan/buddie`; verbatim is buddie's quote module.
+
+- Effect anchors `pr:`, `ci:` (GitHub API) and `run:` (session transcript); output the agent typed itself does not count.
+- Work and mandate claims against a task queue (`buddie.toml`, opt-in).
+- Repo anchors `path#sha256`, `path@commit`, `fact=value`.
+- Receipt v0 (in-toto Statement), CLI, MCP server over stdio, Claude Code hook, plugin and marketplace.
+- Results E009 (numbers) and E014 (Search Arena) added.
+
+## verbatim
+
+Each version was measured on reports it was not tuned on. Experiment folders are in `results/`.
+
+## 0.3.3 (2026-10-02)
+
+- Footnote markers inside quotation marks (`[1](url)`, `[2]`) are dropped before search.
+- A truncated response body (`IncompleteRead`) is a failed fetch, not a crash of `check`.
+- E014 in-sample recount: false `NOT_FOUND` 26 of 423 instead of 29; no agent error became a find.
+
+## 0.3.2 (2026-10-02)
+
+- Two replies of one speaker joined across the page's speech tag are found (`note: "speech_tag"`).
+- The page's `[A]t` brackets and `___` blanks are normalized. Mojibake in the text layer is repaired before search.
+- E012: none of 200 labelled agent errors became a find.
+
+## 0.3.1
+
+- New hints that mark an example before the quote marks: `like “`, `something like: “`, `might include: “`,
+  `questions (“`. E011: 2 hits in 249 checked quotes, both correct.
+
+## 0.3.0
+
+- `NOT_A_QUOTE` verdict for words the report itself shows are not a quotation (`boilerplate`, `title`, `example`).
+- Punctuation between words is ignored at the last normalization level.
+- E010: false `NOT_FOUND` 5 of 147 (3.4 %).
+
+## 0.2
+
+- A quote is checked against its own link. A find under another link of the paragraph is `UNCERTAIN`.
+- `UNCERTAIN` doubts and the access cascade (`--access`: browser, Wayback / Common Crawl, open access, copies).
+- PDF via `pdftotext`, compressed bodies, cookie and bot gates are reported as `SOURCE_UNAVAILABLE`.
+
+## 0.1
+
+- `snap`, `text`, `check`: the content-addressed store, the visible-text layer and the quote search.

buddie-1.0.0/LICENSE

@@ -0,0 +1,177 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://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

buddie-1.0.0/MANIFEST.in

@@ -0,0 +1,11 @@
+# The sdist is the package plus what a reader of it needs; experiment data stays in the repository.
+include README.md CHANGELOG.md LICENSE NOTICE PRIVACY.md run.py
+prune results
+prune tests
+prune examples
+prune docs
+prune scripts
+prune assets
+prune hooks
+prune skills
+global-exclude __pycache__ *.py[cod]

buddie-1.0.0/NOTICE

@@ -0,0 +1,21 @@
+buddie (including its quote module verbatim)
+Copyright 2026 tsyyan
+
+The code, documentation, labels, verdicts and comparison files in this repository are licensed under the Apache
+License, Version 2.0 (LICENSE).
+
+Third-party material under results/:
+
+- DeepResearch Bench (muset-ai/DeepResearch-Bench-Dataset; github.com/Ayanami0730/deep_research_bench), Apache-2.0.
+  Report excerpts in results/E004-*, E005-*, E006-*, E009-*, E011-*, E012-*. Du et al., "DeepResearch Bench: A Comprehensive
+  Benchmark for Deep Research Agents", arXiv:2506.11763.
+- DeepResearch Bench II (muset-ai/DeepResearch-Bench-II-Dataset, revision 73c8b00), Apache-2.0. Report excerpts in
+  results/E008-*.
+- ydc-deep-research-evals (SaltyDuck0/ydc-deep-research-evals, revision 59ee56c): no license is stated, so no report
+  text is included. results/E010-* holds only item ids, labels, verdicts, URLs and SHA-256 hashes of the dataset files.
+- Search Arena 24k (lmarena-ai/search-arena-24k, revision fac8dcf): prompts CC-BY-4.0, model answers under the
+  answering providers' terms, so no answer text is included in data files. results/E014-* holds item ids, labels,
+  verdicts, URLs and hashes; its report (REPORT.ru.md) quotes a few short phrases for research and criticism.
+
+Short passages of cited web pages appear in verdict and label files (the closest passage found, a labeller's evidence)
+for the purpose of research and criticism. Full page snapshots are not redistributed.