Repository navigation
Publish verified native foundation and scoped OmniRoute installation - #15
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b60715a37f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| npm rebuild --global --prefix "$OMNIROUTE_PREFIX" \ | ||
| --allow-scripts=omniroute,keytar,onnxruntime-node,tls-client-node,@parcel/watcher,@swc/core,protobufjs,koffi,esbuild \ | ||
| --foreground-scripts --no-audit --no-fund |
There was a problem hiding this comment.
On the documented Node 24/npm 11.4.2 environment, npm rebuild --help supports --ignore-scripts and --foreground-scripts but not --allow-scripts; running this command reports Unknown cli config "--allow-scripts" and still exits successfully. Consequently, a new-host install silently rebuilds without the claimed package allowlist, so lifecycle scripts from dependencies outside this list may run and the recorded “scoped dependency postinstall” guarantee is false. Use an npm-supported mechanism or remove the misleading restriction.
AGENTS.md reference: AGENTS.md:L40-L40
Useful? React with 👍 / 👎.
| "total_rows": 24, | ||
| "native_savings_counter_tools": 5, | ||
| "core_without_native_savings_counter": 10, | ||
| "runtime": 1, |
There was a problem hiding this comment.
Adding the runtime row makes this document contain 24 rows and the stack contain 66 components, but the top-level scope still describes only 14 core plus 9 observation rows and says the Setup view has 65 components. build_ecosystem.py copies this scope into the generated data and template.html renders it verbatim, so the consolidated offline guide publishes coverage that contradicts both these updated counts and its actual contents.
AGENTS.md reference: AGENTS.md:L33-L37
Useful? React with 👍 / 👎.
| "repomix", | ||
| "toon", |
There was a problem hiding this comment.
Neither the memory_rag, omniroute, native_client_acceptance, nor native_counter_snapshot sections record a Repomix or TOON command, result, counter, or comparison, yet listing both component IDs causes manifests/stack.json, the saturation matrix, and the generated explorer to present this receipt as fresh functional evidence for them. Remove these two associations or publish the scoped Repomix/TOON evidence that supports them.
Useful? React with 👍 / 👎.
| "statistics_note": "No upstream lifetime/session savings counter. Listed status or per-conversion commands, if any, are not cumulative avoided-token accounting." | ||
| }, | ||
| "returned_result_summary": "2026-09-20 loaded Desktop get_current_config: active agent-lab project, language server ready, Bash. Initial pattern query returned empty; corrected explicit source request returned '172:def parse_codex(stream, env):'. Earlier native Codex find_symbol/find_referencing_symbols calls succeeded. Bash readiness is not proof of Python semantic support.", | ||
| "returned_result_summary": "2026-09-20 loaded Desktop get_current_config: active agent-lab project, language server ready, Bash. Initial pattern query returned empty; corrected explicit source request returned '172:def parse_codex(stream, env):'. Earlier native Codex find_symbol/find_referencing_symbols calls succeeded. Bash readiness is not proof of Python semantic support. Foundation repair enabled Python/TypeScript alongside Bash. Eight fresh Codex/Claude MCP-context source/reference checks passed for Python and CJS.", |
There was a problem hiding this comment.
This update now says eight fresh Codex/Claude Serena checks passed and adds the foundation receipt as evidence, but the same row's native_client_evidence.claude still says that no separate current Claude semantic result was established. Machine consumers therefore receive mutually exclusive acceptance states for the same component, and the generated explorer shows the new pass without preserving the stale limitation. Update the per-client evidence to describe the fresh Claude scope.
AGENTS.md reference: AGENTS.md:L17-L17
Useful? React with 👍 / 👎.
| "observed_at": "2026-09-20T15:50:40.270638+00:00", | ||
| "scope": "Default index-root bytes/4 ledger, generation 2; includes repeated verification retrievals. The isolated clean-prefix ledger is excluded. This is a whole-file counterfactual estimate, not measured provider savings.", | ||
| "summary": "39,760 estimated tokens saved at 2026-09-20 14:55:58 UTC in the persistent default index ledger, including validation retrievals. Earlier dated snapshot: 29,820. Isolated clean-prefix totals remain separate; schema-size estimates are never added.", | ||
| "summary": "Foundation snapshot:49,700 retained upstream whole-file bytes/4 estimate (earlier39,760); includes validation and both fresh native reads. Shared ledger, not additive with session subsets.", |
There was a problem hiding this comment.
The lifetime snapshot is now 49,700 and correctly moves 39,760 into earlier_observations, but this row's baseline_summary still calls 39,760 “the retained ... estimate” without its earlier timestamp. The explorer presents that baseline detail next to the new lifetime summary, leaving two different values described as the retained estimate; update it to 49,700 or explicitly label 39,760 as the earlier snapshot.
AGENTS.md reference: AGENTS.md:L19-L22
Useful? React with 👍 / 👎.
The catalog lacked a portable foundation guide and the installed OmniRoute gateway. This adds the upstream setup recipes, registers OmniRoute as optional, and embeds the new foundation evidence in the HTML and token manifest: 66 selected components, 512 repository identities and 83 receipts.
The receipt records the repaired Python/JavaScript retrieval configuration, seven successful MCP calls in each native client, same-process jCodeMunch estimates, and native usage matched with observation. It also verifies the clean pinned OmniRoute install and scoped lifecycle. Failed compression fidelity, gateway quota/model/argument failures, and unsupported native encrypted restore remain explicit; no successful gateway inference or causal provider saving is claimed.
Validation: 554 tests passed with 40 environment-dependent skips; manifest, catalog, convergence and HTML rebuild checks passed. Independent review verified the raw native/observer values, hashes, redaction and counter provenance. Desktop/mobile HTML checks found no overflow or external resource requests; redacted secret scan passed. The final HTML source links are pinned to the verified source commit.