execweave 0.6.9__tar.gz → 0.7.0__tar.gz

{execweave-0.6.9 → execweave-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: execweave
-Version: 0.6.9
+Version: 0.7.0
 Summary: Turn AI-agent runtime behavior into a local interactive execution graph.
 Project-URL: Homepage, https://github.com/Irish-kw/ExecWeave
 Project-URL: Repository, https://github.com/Irish-kw/ExecWeave
@@ -66,7 +66,7 @@ Description-Content-Type: text/markdown
 
 **See what AI agents actually do on your machine.**
 
-ExecWeave is a source-available, local-first observability project that turns AI-agent activity into an interactive execution graph while keeping observed evidence, provider content, and derived inference explicitly separated. Starting with v0.6.8, the project is licensed for noncommercial use under PolyForm Noncommercial 1.0.0.
+ExecWeave is a source-available, local-first observability project that turns AI-agent activity into an interactive execution graph while keeping observed evidence, provider content, and derived inference explicitly separated.
 
 > **Event is ground truth. The graph is a materialized view.**
 
@@ -82,11 +82,7 @@ Install the latest published wheel/sdist from PyPI:
 python -m pip install -U execweave
 ```
 
-The package version on `main` is currently **v0.6.9**. The published release may lag main; test the exact mainline build with:
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+The current release is **v0.7.0**.
 
 For development:
 
@@ -122,6 +118,10 @@ execweave record --open -- python my_agent.py
 
 `execweave top -- codex` keeps the Agent interactive in the launch terminal while opening/attaching the detached Top dashboard according to the host environment.
 
+v0.7.0 extends ExecWeave from provider/runtime observability into attributable multi-agent execution traces. Where providers expose authoritative identity or lifecycle evidence, ExecWeave records parent/child relationships, delegated tasks, inter-agent messages, waits/results, tool ownership, and provider-supplied reasoning/content without promoting weak timing proximity into causal claims. Current provider coverage includes Claude Code, OpenAI Codex, Cursor, OpenCode, and Google Antigravity; missing or ambiguous evidence causes abstention rather than heuristic linkage.
+
+Conversation evidence is now surfaced directly from each ExecWeave run. Existing provider content is indexed, validated provider transcripts that would otherwise require browsing Agent-specific folders are copied into the run-local SHA-256 content store, and finalized runs generate `conversations.md` plus `conversations.json`. Static and Live dashboards link directly to these run-local records. Live content serving is token-authenticated and restricted to validated run-local conversation indexes and SHA-256 content paths; arbitrary local files and path traversal are not exposed.
+
 ## v0.6.9: full-fidelity observability with explicit evidence boundaries
 
 v0.6.9 extends provider/runtime observability beyond compact metadata. When a supported integration point explicitly supplies content, ExecWeave can preserve the **complete supplied value** in a local SHA-256 content-addressed store while keeping only a reference in the semantic event stream.
@@ -282,7 +282,7 @@ Portable filesystem observation is session-correlated rather than process-causal
 
 ## Performance and large-run safety
 
-v0.6.3 added bounded filesystem/viewer protections, incremental Live JSONL tailing, and large-graph safety guards. v0.6.4 added detached Top and unified provisional live sidecars for configured provider integrations. These remain in v0.6.9; the project has **not** migrated Live to SSE, artifact storage to SQLite, the renderer to Canvas/WebGL, or collectors to Rust solely for this release.
+ExecWeave includes bounded filesystem/viewer protections, incremental Live JSONL tailing, large-graph safety guards, detached Top, and provisional live sidecars for configured provider integrations.
 
 The reproducible incremental `GraphAccumulator` reference result reaches **164,273 ev/s** at 1M synthetic events on the documented GitHub Actions workload. This is a graph-accumulation benchmark, not end-to-end collector/browser throughput.
 
@@ -306,6 +306,8 @@ A provider-integrated run may contain:
 ├── viewer.html
 ├── semantic.jsonl
 ├── content/sha256/...
+├── conversations.md
+├── conversations.json
 ├── events.semantic.jsonl
 ├── graph.semantic.json
 ├── viewer.semantic.html
@@ -319,15 +321,13 @@ Derived correlation never rewrites the raw runtime or provider sidecar evidence.
 
 ## Privacy
 
-ExecWeave is local-first: captures, content blobs, graphs, reports, and viewers remain local by default. The **OS runtime collector** does not intentionally capture file contents or raw read/write byte buffers. That boundary must not be confused with the v0.6.9 **provider full-fidelity content store**: supported hooks/APIs can explicitly supply prompts, tool arguments/results, model responses, reasoning/thinking text, shell output, file content, or other sensitive values, and ExecWeave can preserve those values completely.
+ExecWeave is local-first: captures, content blobs, graphs, reports, and viewers remain local by default. The **OS runtime collector** does not intentionally capture file contents or raw read/write byte buffers. That boundary must not be confused with the **provider full-fidelity content store introduced in v0.6.9**: supported hooks/APIs can explicitly supply prompts, tool arguments/results, model responses, reasoning/thinking text, shell output, file content, or other sensitive values, and ExecWeave can preserve those values completely.
 
 Do not assume content has been secret-redacted. Commands, paths, endpoint metadata, identifiers, model metadata, prompts, tool values, and content blobs can all be sensitive. Review the entire run directory before sharing it.
 
 ## Current status
 
-ExecWeave `main` is currently **v0.6.9** and under active release hardening. The latest published package/release can lag main until a GitHub Release is explicitly published; the publish workflow verifies that the release tag exactly matches the package version before PyPI upload.
-
-v0.6.9 combines cross-platform runtime collection, materialized execution graphs, standalone/live viewing, conservative provider↔runtime correlation, full-fidelity content-addressed provider evidence, evidence grades, bounded rule packs, an explicit runtime threat/fidelity contract, and honest local run-integrity sealing. Observed evidence and inference remain separate by design.
+v0.7.0 combines cross-platform runtime collection, materialized execution graphs, standalone/live viewing, conservative provider↔runtime correlation, full-fidelity content-addressed provider evidence, attributable multi-agent execution traces, direct run-local conversation access, evidence grades, bounded rule packs, an explicit runtime threat/fidelity contract, and honest local run-integrity sealing. Observed evidence and inference remain separate by design.
 
 ## Documentation
 
@@ -355,4 +355,4 @@ Contributions are welcome, especially around native OS collectors, Agent/IDE ada
 
 ## License
 
-ExecWeave v0.6.8 and later are licensed under the **PolyForm Noncommercial License 1.0.0**. Noncommercial use, modification, and redistribution are permitted under those terms. Commercial use requires a separate written commercial license from the licensor. Earlier versions already released under MIT remain under the terms that accompanied those versions. See [`LICENSE`](LICENSE).
+Starting with v0.6.8, ExecWeave is licensed under the **PolyForm Noncommercial License 1.0.0**. Noncommercial use, modification, and redistribution are permitted under its terms. Commercial use requires a separate written commercial license from the licensor. See [`LICENSE`](LICENSE).

{execweave-0.6.9 → execweave-0.7.0}/README.de.md

@@ -15,7 +15,7 @@
 
 **Sehen Sie, was KI-Agenten auf Ihrem Rechner tatsächlich tun.**
 
-ExecWeave ist ein source-available, local-first Observability-Projekt, das Aktivitäten von KI-Agenten in einen interaktiven Execution Graph überführt und dabei observed evidence, provider content und derived inference ausdrücklich getrennt hält. Ab v0.6.8 gilt die PolyForm Noncommercial License 1.0.0; kommerzielle Nutzung ist nicht erlaubt.
+ExecWeave ist ein source-available, local-first Observability-Projekt, das Aktivitäten von AI Agents in einen interaktiven execution graph umwandelt und observed evidence, provider content sowie derived inference ausdrücklich getrennt hält.
 
 > **Das Event ist die Ground Truth. Der Graph ist eine materialized view.**
 
@@ -31,11 +31,7 @@ Installieren Sie das neueste veröffentlichte wheel/sdist von PyPI:
 python -m pip install -U execweave
 ```
 
-Die Package-Version auf `main` ist derzeit **v0.6.9**. Die veröffentlichte Release kann hinter main zurückliegen. Um exakt den aktuellen Mainline-Build zu testen:
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+Die aktuelle stabile Version ist **v0.6.9**.
 
 Für die Entwicklung:
 
@@ -228,7 +224,7 @@ Portable Filesystem Observation ist session-correlated und nicht process-causal;
 
 ## Performance und Large-run Safety
 
-v0.6.3 brachte begrenzte Filesystem-/Viewer-Schutzmaßnahmen, inkrementelles Live-JSONL-Tailing und Large-graph Safety Guards. v0.6.4 ergänzte detached Top sowie einen gemeinsamen provisional live sidecar für konfigurierte Provider-Integrationen. Diese Fähigkeiten bleiben in v0.6.9 erhalten. Für diese Release wurde Live **nicht** allein aus Architekturgründen auf SSE, Artifact Storage auf SQLite, der Renderer auf Canvas/WebGL oder die Collectors auf Rust migriert.
+ExecWeave umfasst begrenzte filesystem/viewer protections, inkrementelles Live-JSONL-Tailing, large-graph safety guards, detached Top und provisorische Live-Sidecars für konfigurierte Provider-Integrationen.
 
 Das reproduzierbare inkrementelle `GraphAccumulator`-Referenzergebnis erreicht bei 1M synthetischen Events auf dem dokumentierten GitHub-Actions-Workload **164,273 ev/s**. Dies ist ein Graph-Accumulation-Benchmark und kein End-to-end Collector-/Browser-Throughput.
 
@@ -269,8 +265,6 @@ Gehen Sie nicht davon aus, dass Content secret-redacted ist. Commands, Paths, En
 
 ## Aktueller Status
 
-ExecWeave `main` ist derzeit **v0.6.9** und befindet sich im Release Hardening. Das neueste öffentliche Package/Release kann hinter main liegen, bis eine GitHub Release ausdrücklich veröffentlicht wird. Der Publish Workflow prüft vor dem PyPI Upload, dass Release Tag und Package Version exakt übereinstimmen.
-
 v0.6.9 kombiniert cross-platform runtime collection, materialisierte Execution Graphs, standalone/live Viewing, konservative Provider↔Runtime-Korrelation, content-addressed full-fidelity Provider Evidence, Evidence Grades, begrenzte Rule Packs, einen expliziten Runtime Threat/Fidelity Contract und ehrliches lokales Run-integrity Sealing. Observed evidence und inference bleiben per Design getrennt.
 
 ## Dokumentation
@@ -299,4 +293,4 @@ Beiträge sind willkommen, besonders zu nativen OS Collectors, Agent-/IDE-Adapte
 
 ## License
 
-ExecWeave v0.6.8 und spätere Versionen stehen unter der **PolyForm Noncommercial License 1.0.0**. Nichtkommerzielle Nutzung, Änderung und Weitergabe sind unter diesen Bedingungen erlaubt; jede kommerzielle Nutzung erfordert eine separate schriftliche kommerzielle Lizenz. Bereits zuvor unter MIT veröffentlichte Versionen behalten ihre damaligen Lizenzbedingungen. Siehe [`LICENSE`](LICENSE).
+ExecWeave steht unter der **PolyForm Noncommercial License 1.0.0**. Nichtkommerzielle Nutzung, Änderung und Weiterverteilung sind gemäß den Lizenzbedingungen gestattet. Kommerzielle Nutzung erfordert eine separate schriftliche kommerzielle Lizenz des Lizenzgebers. Siehe [`LICENSE`](LICENSE).

{execweave-0.6.9 → execweave-0.7.0}/README.fr.md

@@ -15,7 +15,7 @@
 
 **Voyez ce que les agents IA font réellement sur votre machine.**
 
-ExecWeave est un projet source-available, local-first, d’observabilité qui transforme l’activité des agents IA en graphe d’exécution interactif tout en séparant explicitement les preuves observées, le contenu fourni par les providers et les inférences dérivées. À partir de v0.6.8, le projet est sous PolyForm Noncommercial 1.0.0 et l’usage commercial n’est pas autorisé.
+ExecWeave est un projet d’observabilité source-available et local-first qui transforme l’activité des agents IA en execution graph interactif, tout en séparant explicitement observed evidence, provider content et derived inference.
 
 > **L’événement est la ground truth. Le graphe est une materialized view.**
 
@@ -31,11 +31,7 @@ Installez la dernière wheel/sdist publiée sur PyPI :
 python -m pip install -U execweave
 ```
 
-La version du package sur `main` est actuellement **v0.6.9**. La release publiée peut être en retard sur main ; pour tester exactement le mainline courant :
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+La version stable actuelle est **v0.6.9**.
 
 Pour le développement :
 
@@ -228,7 +224,7 @@ L’observation filesystem portable est session-correlated et non process-causal
 
 ## Performance et sécurité des grands runs
 
-v0.6.3 a ajouté des protections filesystem/viewer bornées, le tail Live JSONL incrémental et des garde-fous pour les grands graphes. v0.6.4 a ajouté Top détaché et le sidecar live provisoire partagé pour les intégrations provider configurées. Ces capacités restent en v0.6.9. Cette release n’a **pas** migré Live vers SSE, le stockage d’artefacts vers SQLite, le renderer vers Canvas/WebGL, ni les collectors vers Rust simplement pour changer d’architecture.
+ExecWeave inclut des protections bornées pour le filesystem/viewer, le suivi incrémental Live JSONL, des garde-fous pour les grands graphes, Top détaché et des live sidecars provisoires pour les provider integrations configurées.
 
 Le résultat de référence reproductible de l’`GraphAccumulator` incrémental atteint **164,273 ev/s** à 1M d’événements synthétiques sur le workload GitHub Actions documenté. Il s’agit d’un benchmark d’accumulation de graphe, pas d’un débit end-to-end collector/browser.
 
@@ -269,8 +265,6 @@ Ne supposez pas que le contenu a été secret-redacted. Commandes, chemins, endp
 
 ## État actuel
 
-ExecWeave `main` est actuellement en **v0.6.9** et en phase de release hardening. Le dernier package/release public peut être en retard sur main jusqu’à la publication explicite d’une GitHub Release ; le workflow de publication vérifie que le release tag correspond exactement à la package version avant upload sur PyPI.
-
 v0.6.9 combine collection runtime cross-platform, graphes d’exécution matérialisés, viewers standalone/live, corrélation provider↔runtime conservatrice, preuves provider full-fidelity adressées par contenu, evidence grades, rule packs bornés, contrat explicite runtime threat/fidelity et scellement local d’intégrité avec une frontière de confiance honnête. Les preuves observées et l’inférence restent séparées par conception.
 
 ## Documentation
@@ -299,4 +293,4 @@ Les contributions sont bienvenues, notamment pour les collectors OS natifs, adap
 
 ## License
 
-ExecWeave v0.6.8 et les versions ultérieures sont sous **PolyForm Noncommercial License 1.0.0**. L’usage, la modification et la redistribution non commerciaux sont permis selon ces termes ; tout usage commercial nécessite une licence commerciale écrite distincte. Les versions antérieures déjà publiées sous MIT restent soumises aux conditions qui les accompagnaient. Voir [`LICENSE`](LICENSE).
+ExecWeave est distribué sous **PolyForm Noncommercial License 1.0.0**. L’utilisation, la modification et la redistribution non commerciales sont autorisées selon ses conditions. Toute utilisation commerciale nécessite une licence commerciale écrite distincte du concédant. Voir [`LICENSE`](LICENSE).

{execweave-0.6.9 → execweave-0.7.0}/README.ja.md

@@ -15,7 +15,7 @@
 
 **AI Agent があなたのマシン上で実際に何をしているかを可視化します。**
 
-ExecWeave は source-available、local-first の observability プロジェクトで、AI Agent の活動をインタラクティブな execution graph に変換し、observed evidence、provider content、derived inference を明確に分離します。v0.6.8 以降は PolyForm Noncommercial 1.0.0 の下で提供され、商用利用は許可されません。
+ExecWeave は source-available、local-first の observability プロジェクトで、AI Agent の活動をインタラクティブな execution graph に変換し、observed evidence、provider content、derived inference を明確に分離します。
 
 > **Event が ground truth であり、Graph は materialized view です。**
 
@@ -31,11 +31,7 @@ PyPI から最新の公開 wheel/sdist をインストールします。
 python -m pip install -U execweave
 ```
 
-現在の `main` の package version は **v0.6.9** です。公開 release が main より遅れる場合があります。現在の mainline を直接試すには：
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+現在の正式リリースは **v0.6.9** です。
 
 開発用：
 
@@ -228,7 +224,7 @@ Portable filesystem observation は session-correlated であり process-causal
 
 ## Performance と large-run safety
 
-v0.6.3 では bounded filesystem/viewer protection、incremental Live JSONL tailing、large-graph safety guard を追加し、v0.6.4 では detached Top と configured provider integration 用 provisional live sidecar を追加しました。これらは v0.6.9 にも残っています。本 release だけを理由に Live を SSE、artifact storage を SQLite、renderer を Canvas/WebGL、collector を Rust へ移行してはいません。
+ExecWeave は bounded filesystem/viewer protection、incremental Live JSONL tailing、large-graph safety guard、detached Top、configured provider integration 用 provisional live sidecar を備えています。
 
 再現可能な incremental `GraphAccumulator` reference result は、文書化された GitHub Actions workload の 1M synthetic events で **164,273 ev/s** です。これは graph accumulation benchmark であり、end-to-end collector/browser throughput ではありません。
 
@@ -269,8 +265,6 @@ Content が secret-redacted 済みだと仮定しないでください。Command
 
 ## 現在の状態
 
-ExecWeave `main` は現在 **v0.6.9** で release hardening 中です。公開 package/release は main より遅れる場合があります。GitHub Release を明示的に publish した場合のみ publish workflow が動作し、PyPI upload 前に release tag と package version が完全一致することを検証します。
-
 v0.6.9 は cross-platform runtime collection、materialized execution graph、standalone/live viewer、保守的 provider↔runtime correlation、content-addressed full-fidelity provider evidence、evidence grades、bounded rule packs、明示的 runtime threat/fidelity contract、honest local run-integrity sealing を組み合わせています。Observed evidence と inference は設計上分離されています。
 
 ## ドキュメント
@@ -299,4 +293,4 @@ v0.6.9 は cross-platform runtime collection、materialized execution graph、st
 
 ## License
 
-ExecWeave v0.6.8 以降は **PolyForm Noncommercial License 1.0.0** の下で提供されます。非商用の利用・変更・再配布はその条件に従って許可されますが、商用利用には別途書面による商用ライセンスが必要です。以前に MIT で公開済みの旧バージョンは当時のライセンス条件のままです。詳しくは [`LICENSE`](LICENSE) を参照してください。
+ExecWeave は **PolyForm Noncommercial License 1.0.0** の下で提供されます。非商用の利用・変更・再配布はその条件に従って許可され、商用利用にはライセンサーから別途書面による商用ライセンスが必要です。詳しくは [`LICENSE`](LICENSE) を参照してください。

{execweave-0.6.9 → execweave-0.7.0}/README.ko.md

@@ -15,7 +15,7 @@
 
 **AI Agent가 내 컴퓨터에서 실제로 무엇을 하는지 확인하세요.**
 
-ExecWeave는 AI Agent 활동을 인터랙티브 execution graph로 변환하면서 observed evidence, provider content, derived inference를 명확히 분리하는 source-available, local-first observability 프로젝트입니다. v0.6.8부터 PolyForm Noncommercial 1.0.0이 적용되며 상업적 사용은 허용되지 않습니다.
+ExecWeave는 AI Agent 활동을 인터랙티브 execution graph로 변환하면서 observed evidence, provider content, derived inference를 명확히 분리하는 source-available, local-first observability 프로젝트입니다.
 
 > **Event가 ground truth이고, Graph는 materialized view입니다.**
 
@@ -31,11 +31,7 @@ PyPI에서 최신 공개 wheel/sdist를 설치합니다.
 python -m pip install -U execweave
 ```
 
-현재 `main`의 package version은 **v0.6.9**입니다. 공개 release가 main보다 늦을 수 있습니다. 현재 mainline을 직접 테스트하려면:
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+현재 정식 릴리스는 **v0.6.9**입니다.
 
 개발 설치:
 
@@ -228,7 +224,7 @@ Portable filesystem observation은 session-correlated이며 process-causal이
 
 ## Performance와 large-run safety
 
-v0.6.3은 bounded filesystem/viewer protection, incremental Live JSONL tailing, large-graph safety guard를 추가했고 v0.6.4는 detached Top과 configured provider integration용 provisional live sidecar를 추가했습니다. 이 기능들은 v0.6.9에도 유지됩니다. 이번 release만을 위해 Live를 SSE로, artifact storage를 SQLite로, renderer를 Canvas/WebGL로, collector를 Rust로 전환하지 않았습니다.
+ExecWeave는 bounded filesystem/viewer protection, incremental Live JSONL tailing, large-graph safety guard, detached Top, configured provider integration용 provisional live sidecar를 포함합니다.
 
 재현 가능한 incremental `GraphAccumulator` reference result는 문서화된 GitHub Actions workload에서 1M synthetic events 기준 **164,273 ev/s**입니다. 이는 graph accumulation benchmark이며 end-to-end collector/browser throughput이 아닙니다.
 
@@ -269,8 +265,6 @@ Content가 secret-redacted되었다고 가정하지 마세요. Command, path, en
 
 ## 현재 상태
 
-ExecWeave `main`은 현재 **v0.6.9**이며 release hardening 중입니다. 공개 package/release는 main보다 늦을 수 있습니다. GitHub Release가 명시적으로 publish될 때만 publish workflow가 실행되며 PyPI upload 전에 release tag와 package version이 정확히 일치하는지 검증합니다.
-
 v0.6.9는 cross-platform runtime collection, materialized execution graph, standalone/live viewer, 보수적인 provider↔runtime correlation, content-addressed full-fidelity provider evidence, evidence grades, bounded rule packs, 명시적 runtime threat/fidelity contract, honest local run-integrity sealing을 결합합니다. Observed evidence와 inference는 설계상 분리됩니다.
 
 ## 문서
@@ -299,4 +293,4 @@ v0.6.9는 cross-platform runtime collection, materialized execution graph, stand
 
 ## License
 
-ExecWeave v0.6.8 이상은 **PolyForm Noncommercial License 1.0.0**에 따라 제공됩니다. 비상업적 사용, 수정, 재배포는 해당 조건에 따라 허용되지만 상업적 사용에는 별도의 서면 상업 라이선스가 필요합니다. 이전에 MIT로 공개된 버전은 당시 함께 제공된 라이선스 조건을 유지합니다. 자세한 내용은 [`LICENSE`](LICENSE)를 참고하세요.
+ExecWeave는 **PolyForm Noncommercial License 1.0.0**에 따라 제공됩니다. 비상업적 사용, 수정 및 재배포는 해당 조건에 따라 허용되며, 상업적 사용에는 라이선서의 별도 서면 상업용 라이선스가 필요합니다. [`LICENSE`](LICENSE)를 참조하세요.

{execweave-0.6.9 → execweave-0.7.0}/README.md

@@ -15,7 +15,7 @@
 
 **See what AI agents actually do on your machine.**
 
-ExecWeave is a source-available, local-first observability project that turns AI-agent activity into an interactive execution graph while keeping observed evidence, provider content, and derived inference explicitly separated. Starting with v0.6.8, the project is licensed for noncommercial use under PolyForm Noncommercial 1.0.0.
+ExecWeave is a source-available, local-first observability project that turns AI-agent activity into an interactive execution graph while keeping observed evidence, provider content, and derived inference explicitly separated.
 
 > **Event is ground truth. The graph is a materialized view.**
 
@@ -31,11 +31,7 @@ Install the latest published wheel/sdist from PyPI:
 python -m pip install -U execweave
 ```
 
-The package version on `main` is currently **v0.6.9**. The published release may lag main; test the exact mainline build with:
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+The current release is **v0.7.0**.
 
 For development:
 
@@ -71,6 +67,10 @@ execweave record --open -- python my_agent.py
 
 `execweave top -- codex` keeps the Agent interactive in the launch terminal while opening/attaching the detached Top dashboard according to the host environment.
 
+v0.7.0 extends ExecWeave from provider/runtime observability into attributable multi-agent execution traces. Where providers expose authoritative identity or lifecycle evidence, ExecWeave records parent/child relationships, delegated tasks, inter-agent messages, waits/results, tool ownership, and provider-supplied reasoning/content without promoting weak timing proximity into causal claims. Current provider coverage includes Claude Code, OpenAI Codex, Cursor, OpenCode, and Google Antigravity; missing or ambiguous evidence causes abstention rather than heuristic linkage.
+
+Conversation evidence is now surfaced directly from each ExecWeave run. Existing provider content is indexed, validated provider transcripts that would otherwise require browsing Agent-specific folders are copied into the run-local SHA-256 content store, and finalized runs generate `conversations.md` plus `conversations.json`. Static and Live dashboards link directly to these run-local records. Live content serving is token-authenticated and restricted to validated run-local conversation indexes and SHA-256 content paths; arbitrary local files and path traversal are not exposed.
+
 ## v0.6.9: full-fidelity observability with explicit evidence boundaries
 
 v0.6.9 extends provider/runtime observability beyond compact metadata. When a supported integration point explicitly supplies content, ExecWeave can preserve the **complete supplied value** in a local SHA-256 content-addressed store while keeping only a reference in the semantic event stream.
@@ -231,7 +231,7 @@ Portable filesystem observation is session-correlated rather than process-causal
 
 ## Performance and large-run safety
 
-v0.6.3 added bounded filesystem/viewer protections, incremental Live JSONL tailing, and large-graph safety guards. v0.6.4 added detached Top and unified provisional live sidecars for configured provider integrations. These remain in v0.6.9; the project has **not** migrated Live to SSE, artifact storage to SQLite, the renderer to Canvas/WebGL, or collectors to Rust solely for this release.
+ExecWeave includes bounded filesystem/viewer protections, incremental Live JSONL tailing, large-graph safety guards, detached Top, and provisional live sidecars for configured provider integrations.
 
 The reproducible incremental `GraphAccumulator` reference result reaches **164,273 ev/s** at 1M synthetic events on the documented GitHub Actions workload. This is a graph-accumulation benchmark, not end-to-end collector/browser throughput.
 
@@ -255,6 +255,8 @@ A provider-integrated run may contain:
 ├── viewer.html
 ├── semantic.jsonl
 ├── content/sha256/...
+├── conversations.md
+├── conversations.json
 ├── events.semantic.jsonl
 ├── graph.semantic.json
 ├── viewer.semantic.html
@@ -268,15 +270,13 @@ Derived correlation never rewrites the raw runtime or provider sidecar evidence.
 
 ## Privacy
 
-ExecWeave is local-first: captures, content blobs, graphs, reports, and viewers remain local by default. The **OS runtime collector** does not intentionally capture file contents or raw read/write byte buffers. That boundary must not be confused with the v0.6.9 **provider full-fidelity content store**: supported hooks/APIs can explicitly supply prompts, tool arguments/results, model responses, reasoning/thinking text, shell output, file content, or other sensitive values, and ExecWeave can preserve those values completely.
+ExecWeave is local-first: captures, content blobs, graphs, reports, and viewers remain local by default. The **OS runtime collector** does not intentionally capture file contents or raw read/write byte buffers. That boundary must not be confused with the **provider full-fidelity content store introduced in v0.6.9**: supported hooks/APIs can explicitly supply prompts, tool arguments/results, model responses, reasoning/thinking text, shell output, file content, or other sensitive values, and ExecWeave can preserve those values completely.
 
 Do not assume content has been secret-redacted. Commands, paths, endpoint metadata, identifiers, model metadata, prompts, tool values, and content blobs can all be sensitive. Review the entire run directory before sharing it.
 
 ## Current status
 
-ExecWeave `main` is currently **v0.6.9** and under active release hardening. The latest published package/release can lag main until a GitHub Release is explicitly published; the publish workflow verifies that the release tag exactly matches the package version before PyPI upload.
-
-v0.6.9 combines cross-platform runtime collection, materialized execution graphs, standalone/live viewing, conservative provider↔runtime correlation, full-fidelity content-addressed provider evidence, evidence grades, bounded rule packs, an explicit runtime threat/fidelity contract, and honest local run-integrity sealing. Observed evidence and inference remain separate by design.
+v0.7.0 combines cross-platform runtime collection, materialized execution graphs, standalone/live viewing, conservative provider↔runtime correlation, full-fidelity content-addressed provider evidence, attributable multi-agent execution traces, direct run-local conversation access, evidence grades, bounded rule packs, an explicit runtime threat/fidelity contract, and honest local run-integrity sealing. Observed evidence and inference remain separate by design.
 
 ## Documentation
 
@@ -304,4 +304,4 @@ Contributions are welcome, especially around native OS collectors, Agent/IDE ada
 
 ## License
 
-ExecWeave v0.6.8 and later are licensed under the **PolyForm Noncommercial License 1.0.0**. Noncommercial use, modification, and redistribution are permitted under those terms. Commercial use requires a separate written commercial license from the licensor. Earlier versions already released under MIT remain under the terms that accompanied those versions. See [`LICENSE`](LICENSE).
+Starting with v0.6.8, ExecWeave is licensed under the **PolyForm Noncommercial License 1.0.0**. Noncommercial use, modification, and redistribution are permitted under its terms. Commercial use requires a separate written commercial license from the licensor. See [`LICENSE`](LICENSE).

{execweave-0.6.9 → execweave-0.7.0}/README.ru.md

@@ -15,7 +15,7 @@
 
 **Посмотрите, что ИИ-агенты действительно делают на вашей машине.**
 
-ExecWeave — source-available, local-first проект observability, который преобразует активность ИИ-агентов в интерактивный execution graph и при этом явно разделяет observed evidence, provider content и derived inference. Начиная с v0.6.8 проект распространяется по PolyForm Noncommercial 1.0.0; коммерческое использование не разрешено.
+ExecWeave — source-available, local-first проект наблюдаемости, который преобразует активность AI Agent в интерактивный execution graph и явно разделяет observed evidence, provider content и derived inference.
 
 > **Event — это ground truth. Graph — materialized view.**
 
@@ -31,11 +31,7 @@ ExecWeave — source-available, local-first проект observability, кото
 python -m pip install -U execweave
 ```
 
-Версия package в `main` сейчас **v0.6.9**. Опубликованный release может отставать от main; чтобы протестировать текущий mainline напрямую:
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+Текущий стабильный релиз — **v0.6.9**.
 
 Для разработки:
 
@@ -228,7 +224,7 @@ Portable filesystem observation является session-correlated, а не pro
 
 ## Performance и large-run safety
 
-v0.6.3 добавил bounded filesystem/viewer protections, incremental Live JSONL tailing и large-graph safety guards. v0.6.4 добавил detached Top и общий provisional live sidecar для настроенных provider integrations. Эти возможности остаются в v0.6.9. Ради этой release проект **не** мигрировал Live на SSE, artifact storage на SQLite, renderer на Canvas/WebGL или collectors на Rust только ради смены архитектуры.
+ExecWeave включает bounded filesystem/viewer protection, incremental Live JSONL tailing, large-graph safety guards, detached Top и provisional live sidecars для настроенных provider integrations.
 
 Воспроизводимый reference result incremental `GraphAccumulator` достигает **164,273 ev/s** на 1M synthetic events в документированном GitHub Actions workload. Это benchmark накопления graph, а не end-to-end collector/browser throughput.
 
@@ -269,8 +265,6 @@ ExecWeave — local-first: captures, content blobs, graphs, reports и viewers
 
 ## Текущий статус
 
-ExecWeave `main` сейчас **v0.6.9** и проходит release hardening. Последний публичный package/release может отставать от main до явной публикации GitHub Release; publish workflow проверяет точное совпадение release tag и package version до загрузки в PyPI.
-
 v0.6.9 объединяет cross-platform runtime collection, materialized execution graphs, standalone/live viewing, консервативную provider↔runtime correlation, content-addressed full-fidelity provider evidence, evidence grades, bounded rule packs, явный runtime threat/fidelity contract и честный local run-integrity seal. Observed evidence и inference остаются разделёнными по дизайну.
 
 ## Документация
@@ -299,4 +293,4 @@ v0.6.9 объединяет cross-platform runtime collection, materialized exec
 
 ## License
 
-ExecWeave v0.6.8 и более поздние версии распространяются по **PolyForm Noncommercial License 1.0.0**. Некоммерческие использование, изменение и распространение разрешены в соответствии с этими условиями; любое коммерческое использование требует отдельной письменной коммерческой лицензии. Ранее выпущенные под MIT версии сохраняют условия лицензии, сопровождавшие их на момент выпуска. См. [`LICENSE`](LICENSE).
+ExecWeave распространяется по **PolyForm Noncommercial License 1.0.0**. Некоммерческое использование, изменение и распространение разрешены на условиях лицензии. Для коммерческого использования требуется отдельная письменная коммерческая лицензия от лицензиара. См. [`LICENSE`](LICENSE).

{execweave-0.6.9 → execweave-0.7.0}/README.zh-CN.md

@@ -15,7 +15,7 @@
 
 **看清 AI Agent 在你的机器上究竟做了什么。**
 
-ExecWeave 是一个 source-available、local-first 的可观测性项目，把 AI Agent 活动转换为交互式 execution graph，并明确区分 observed evidence、provider content 与 derived inference。自 v0.6.8 起采用 PolyForm Noncommercial 1.0.0，仅允许非商业用途。
+ExecWeave 是一个 source-available、local-first 的可观测性项目，把 AI Agent 活动转换为交互式 execution graph，并明确区分 observed evidence、provider content 与 derived inference。
 
 > **Event 是 ground truth；Graph 是 materialized view。**
 
@@ -31,11 +31,7 @@ ExecWeave 是一个 source-available、local-first 的可观测性项目，把 A
 python -m pip install -U execweave
 ```
 
-当前 `main` 的包版本是 **v0.6.9**。正式 release 可能晚于 main；要测试当前 mainline：
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+当前正式版本是 **v0.6.9**。
 
 开发安装：
 
@@ -230,7 +226,7 @@ Portable filesystem observation 是 session-correlated，而不是 process-causa
 
 ## Performance 与 large-run safety
 
-v0.6.3 加入 bounded filesystem/viewer protection、incremental Live JSONL tailing 与 large-graph safety guard；v0.6.4 加入 detached Top，以及 configured provider integration 共用的 provisional live sidecar。这些能力继续保留在 v0.6.9。本次 release **没有**仅为架构替换而把 Live 迁移到 SSE、artifact storage 改成 SQLite、renderer 改成 Canvas/WebGL，或把 collector 改写成 Rust。
+ExecWeave 包含 bounded filesystem/viewer protection、incremental Live JSONL tailing、large-graph safety guard、detached Top，以及 configured provider integration 使用的 provisional live sidecar。
 
 可复现的 incremental `GraphAccumulator` reference result 在文档化 GitHub Actions workload 上，1M synthetic events 达到 **164,273 ev/s**。这是 graph accumulation benchmark，而不是 end-to-end collector/browser throughput。
 
@@ -273,8 +269,6 @@ ExecWeave 是 local-first：capture、content blob、graph、report、viewer 默
 
 ## 当前状态
 
-ExecWeave `main` 当前是 **v0.6.9**，正在进行 release hardening。最新公开 package/release 可能晚于 main；只有明确发布 GitHub Release 才会触发 publish workflow，且 workflow 会先验证 release tag 与 package version 完全一致，再上传 PyPI。
-
 v0.6.9 组合 cross-platform runtime collection、materialized execution graph、standalone/live viewer、保守的 provider↔runtime correlation、content-addressed full-fidelity provider evidence、evidence grades、bounded rule packs、明确的 runtime threat/fidelity contract，以及诚实描述信任边界的 local run-integrity sealing。Observed evidence 与 inference 从设计上保持分离。
 
 ## 文档
@@ -303,4 +297,4 @@ v0.6.9 组合 cross-platform runtime collection、materialized execution graph
 
 ## License
 
-ExecWeave 自 v0.6.8 起采用 **PolyForm Noncommercial License 1.0.0**。依该许可可进行非商业使用、修改和分发；任何商业用途都需要另行取得书面商业许可。此前已按 MIT 发布的旧版本仍维持当时随附的许可条款。详见 [`LICENSE`](LICENSE)。
+ExecWeave 采用 **PolyForm Noncommercial License 1.0.0**。依许可条款可进行非商业使用、修改和分发；商业用途需要另行取得许可方的书面商业许可。详见 [`LICENSE`](LICENSE)。

{execweave-0.6.9 → execweave-0.7.0}/README.zh-TW.md

@@ -15,7 +15,7 @@
 
 **看見 AI Agent 在你的電腦上實際做了什麼。**
 
-ExecWeave 是一個 source-available、local-first 的可觀測性專案，會把 AI Agent 活動轉成互動式 execution graph，並明確區分 observed evidence、provider content 與 derived inference。自 v0.6.8 起採用 PolyForm Noncommercial 1.0.0，僅允許非商業用途。
+ExecWeave 是一個 source-available、local-first 的可觀測性專案，會把 AI Agent 活動轉成互動式 execution graph，並明確區分 observed evidence、provider content 與 derived inference。
 
 > **Event 是 ground truth；Graph 是 materialized view。**
 
@@ -31,11 +31,7 @@ ExecWeave 是一個 source-available、local-first 的可觀測性專案，會
 python -m pip install -U execweave
 ```
 
-目前 `main` 的套件版本是 **v0.6.9**。正式 release 可能晚於 main；若要測試目前 mainline：
-
-```bash
-python -m pip install --upgrade --force-reinstall "execweave @ git+https://github.com/Irish-kw/ExecWeave.git@main"
-```
+目前正式版本是 **v0.6.9**。
 
 開發安裝：
 
@@ -230,7 +226,7 @@ Portable filesystem observation 是 session-correlated，不是 process-causal
 
 ## Performance 與 large-run safety
 
-v0.6.3 加入 bounded filesystem/viewer protection、incremental Live JSONL tailing 與 large-graph safety guard；v0.6.4 加入 detached Top，以及 configured provider integration 共用的 provisional live sidecar。這些能力都保留在 v0.6.9。這次 release **沒有**僅為架構替換而把 Live 遷移到 SSE、artifact storage 改成 SQLite、renderer 改成 Canvas/WebGL，或把 collector 改寫成 Rust。
+ExecWeave 具備 bounded filesystem/viewer protection、incremental Live JSONL tailing、large-graph safety guard、detached Top，以及 configured provider integration 使用的 provisional live sidecar。
 
 可重現的 incremental `GraphAccumulator` reference result 在文件化的 GitHub Actions workload 上，1M synthetic events 達到 **164,273 ev/s**。這是 graph accumulation benchmark，不是 end-to-end collector/browser throughput。
 
@@ -273,8 +269,6 @@ ExecWeave 是 local-first：capture、content blob、graph、report、viewer 預
 
 ## 目前狀態
 
-ExecWeave `main` 目前是 **v0.6.9**，正在進行 release hardening。最新公開 package/release 可能會晚於 main；只有明確發布 GitHub Release 才會觸發 publish workflow，而且 workflow 會先驗證 release tag 與 package version 完全一致再上傳 PyPI。
-
 v0.6.9 整合 cross-platform runtime collection、materialized execution graph、standalone/live viewer、保守的 provider↔runtime correlation、content-addressed full-fidelity provider evidence、evidence grades、bounded rule packs、明確的 runtime threat/fidelity contract，以及誠實定義信任邊界的 local run-integrity sealing。Observed evidence 與 inference 仍從設計上分離。
 
 ## 文件
@@ -303,4 +297,4 @@ v0.6.9 整合 cross-platform runtime collection、materialized execution graph
 
 ## License
 
-ExecWeave v0.6.8 起採用 **PolyForm Noncommercial License 1.0.0**。依該授權可作非商業使用、修改與散布；任何商業用途都需要另外取得書面商業授權。先前已依 MIT 發布的舊版本仍維持當時隨附的授權條款。詳見 [`LICENSE`](LICENSE)。
+ExecWeave 採用 **PolyForm Noncommercial License 1.0.0**。依授權條款可進行非商業使用、修改與散布；商業用途需要另外取得授權方的書面商業授權。詳見 [`LICENSE`](LICENSE)。

execweave-0.7.0/docs/antigravity-hooks.md

@@ -0,0 +1,78 @@
+# Google Antigravity hooks
+
+ExecWeave supports the current Google Antigravity CLI through its native hooks and OS-runtime collection. The Antigravity CLI executable is `agy`.
+
+```bash
+execweave live --open -- antigravity
+# ExecWeave resolves this friendly name to the official `agy` CLI binary.
+```
+
+On the first provider-integrated run, Antigravity may ask whether the ExecWeave hook should be allowed. Choose **Allow / Yes** if you want provider-level tool and invocation evidence in addition to OS-runtime telemetry.
+
+## Automatic bootstrap
+
+For `agy` / `antigravity`, ExecWeave conservatively merges one named hook definition into:
+
+```text
+~/.gemini/config/hooks.json
+```
+
+The generated configuration can be inspected without writing it:
+
+```bash
+execweave-antigravity-hook --print-config
+```
+
+ExecWeave deliberately uses passive observation points:
+
+- `PostToolUse` for tool-call/result evidence.
+- `PreInvocation` and `PostInvocation` for invocation lifecycle metadata.
+- **No `PreToolUse` hook is installed**, because that hook participates in Antigravity's permission decision and ExecWeave must not approve, deny, or alter a user's tool permission flow.
+
+Existing unrelated named hooks are preserved. ExecWeave refuses to replace an existing hook definition with the same name rather than silently overwriting it.
+
+## Layered recorder
+
+```bash
+execweave-antigravity-record --open -- antigravity
+```
+
+The recorder keeps runtime evidence, Antigravity semantic evidence, and derived correlation artifacts separate. Normal `execweave live --open -- antigravity` also auto-bootstraps the hook integration and exposes the live graph.
+
+## Evidence scope
+
+Antigravity evidence has two deliberately different boundaries.
+
+The primary hook evidence is bounded to values actually supplied by the hook contract. ExecWeave can record:
+
+- invocation lifecycle metadata exposed by `PreInvocation` / `PostInvocation`;
+- tool identity, arguments, result/error status exposed by `PostToolUse`;
+- declared command or file targets when those fields are present in the supplied tool arguments;
+- complete hook-supplied values in the local content-addressed store where supported.
+
+For a successful `invoke_subagent`, ExecWeave can additionally use the local `transcriptPath` supplied by the hook as **identity-correlation evidence**. The transcript record layout is treated as a live-verified Antigravity implementation wire, not as a public stable provider contract. A child assignment is emitted only when all conservative checks agree, including:
+
+- the parent transcript is a canonical Antigravity brain transcript for the hook's parent `conversationId` and is read as a complete newline-terminated JSONL snapshot;
+- exactly one matching `PLANNER_RESPONSE` / `invoke_subagent` request is immediately followed by an `INVOKE_SUBAGENT` result record;
+- the transcript request's `Subagents` value exactly matches the successful `PostToolUse` request, and result count/order exactly matches request count/order;
+- every child `conversationId` is non-empty, unique, and different from the parent;
+- every child `logAbsoluteUri` is a canonical local file URI whose Antigravity brain path agrees with that child ID and the same Antigravity application-data root;
+- the expected/inherited workspace is valid, and when the result exposes `workspaceUris`, the expected workspace is present there.
+
+Malformed JSONL, torn writes, ambiguous duplicate matches, count mismatches, non-canonical URIs, workspace mismatches, duplicate/self child IDs, or any other failed validation cause ExecWeave to **abstain** and keep the original request-only evidence. No timestamp proximity is used to bridge transcript records.
+
+When validation succeeds, the graph can contain:
+
+```text
+subtask --ASSIGNED_AGENT_TASK--> child agent
+```
+
+That edge is exact identity evidence (`identity_exact: true`) with identity method `validated_transcript_record_order_and_provider_ids`. It is **not** a child-lifecycle claim. The parent transcript does not by itself create `SPAWNED_AGENT`, returned-result, close, completion, or child-execution state; child provider hooks remain authoritative for those lifecycle observations.
+
+The assignment edge does not copy `transcriptPath` or child `logAbsoluteUri` into its graph attributes. Hook-supplied metadata may still be preserved separately by the full-fidelity content layer according to the normal provider-metadata rules.
+
+This integration does **not** claim access to hidden model state, private chain-of-thought, an unseen provider-side request, or any content not delivered to an observed local surface. OS-runtime evidence remains a separate layer.
+
+## Legacy Gemini CLI
+
+The older `execweave-gemini-*` entry points remain available for compatibility with existing installations and archived workflows. They are not the current Google CLI path documented by ExecWeave; new usage should target Antigravity with `agy`.

{execweave-0.6.9 → execweave-0.7.0}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "hatchling.build"
 
 [project]
 name = "execweave"
-# ExecWeave v0.6.9 release metadata
-version = "0.6.9"
+# ExecWeave v0.7.0 release metadata
+version = "0.7.0"
 description = "Turn AI-agent runtime behavior into a local interactive execution graph."
 readme = "README.md"
 requires-python = ">=3.10"