Repository navigation
Consolidate Classic into Ledger and protect drafts and local APIs - #262
Coding-Dev-Tools wants to merge 35 commits into
Conversation
The dashboard now has one interface. Each capability Classic had was rebuilt into the Ledger view where people look for it, then Classic was removed: - Library: Markdown rendering (DOM-only renderer, never parses HTML) with a Write/Preview editor, unsaved-changes guard and Ctrl/Cmd+S; word-level history diffs; per-memory audit trail; drag and Alt+Arrow ordering; one Import menu (Quick import of files, folders or a server path, document import, code indexing and PostgreSQL schema import). - Ask: per-result retrieval scores. Home: memory health and the selected project's last-session handoff (GET /api/proactive gains `repo`). - Explore: colour by cluster, type or connections, palettes, editable type colours, a legend, an entity and relation explorer, keyboard navigation. - Activity: bi-temporal Timeline cards with a superseded-only filter (replacing the Supersessions tab), receipt-chain status, packed-context accounting. - Settings: workspace Describe, Copy, Merge, Export and Delete; plan status and plan-aware hosted messages; automation run status and proposals. - Everywhere: connection and semantic-search banners, an accessible in-page dialog instead of native confirm/prompt, Solarized and Sepia themes. Removed: engraphis/classic_assets, the duplicate engraphis/static tree, the dashboard switcher, the graph compat adapter, and the Marked and DOMPurify copies. /classic redirects to /. Shipped dashboard assets drop from 3.51 MB in three trees to 1.85 MB in one. Hardening found on the way: the zero-config API refuses requests a browser marks as cross-site unless they passed a CORS preflight; the Markdown renderer is linear on adversarial input; dialogs drop bidi controls; secure erase sends its required confirmation; a stalled graph download no longer blocks Reload; the CSP release gate covers Ledger and every first-party script. Tests, docs, packaging and the public offline evidence artifact are updated to match (fixture results unchanged). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QRXYjdLRbDr9oEUU4qdZ2m
Without ENGRAPHIS_API_TOKEN the dashboard trusts loopback peers. A website that points its own name at 127.0.0.1 (DNS rebinding) becomes same-origin with that name, so its requests came from a loopback peer, passed every Origin check and could read or write the whole /api surface. - The zero-config gate now also checks Host: it accepts exactly `localhost`, IP literals (an address cannot be rebound), the bind host, the ENGRAPHIS_DASHBOARD_URL host and the new ENGRAPHIS_ALLOWED_HOSTS names. `localhost.` and `*.localhost` are refused because some resolvers send them to DNS. Token mode is unchanged. - The check covers the `/dashboard/review` routes as well as `/api`; `/mcp` keeps the SDK's own Host allow-list. - The read-only graph server (engraphis-graph-server) applies the same check when it runs without a token. - The cross-site guard no longer lets a header vouch for another site's page: with ENGRAPHIS_CORS_ORIGINS=* any page could pass the preflight and send the dashboard's header. - Ledger shows the server's reason when a request is refused instead of reporting the server as unreachable. The helpers live in engraphis/netutil.py and are unit-tested; the public offline evidence artifact is re-exported for the new sources (fixture results unchanged). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QRXYjdLRbDr9oEUU4qdZ2m
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: 679dcbb1f9
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 22ce45e0da
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a1af2b2224
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f1f85a4925
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: de8c168b95
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 606d3e579b
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 95999cb42e
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f07db1ff95
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".
Coding-Dev-Tools
changed the title
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7a1b05e8da
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 12102ec86c
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".
| function keep(name) { | ||
| const parts = String(name || '').split('/').filter(Boolean); | ||
| return !parts.slice(0, -1).some(SKIPPED_DIRECTORY); |
There was a problem hiding this comment.
When a user selects a project folder, this predicate checks only ancestor directories, so sensitive leaf files such as a root .env, credentials.json, or id_rsa are uploaded. The folder-path flow has the same exposure because its new default * pattern reaches the server-side walker, which also excludes hidden directories but not sensitive filenames. Content scanning does not close this gap for every credential format—for example, an .env containing an authenticated SMTP URL is accepted and persisted—after which the record can enter exports or shared-workspace sync. Reject sensitive filenames server-side for both upload and path imports rather than relying on content detection.
AGENTS.md reference: AGENTS.md:L248-L249
Useful? React with 👍 / 👎.
Validated local candidate: fc13a0c414b4422b8925d81ab154b464702751ed - fc13a0c414b4422b8925d81ab154b464702751ed docs(readme): pin chart to verified evidence commit - 15a17a0712b411a43e5d4d9ea62f355ca4598d2d docs(evidence): bind offline fixtures to import safety repairs - d6c8252f09334ebb3ddb3df1864c77925017ab5f fix(dashboard): filter credential rc files before upload - 98d48c33d207c66bfa24e74b3974e8cad2e5b123 fix(import): exclude standard credential rc files before processing - d81d0df34c3e285d551c06791311f13dd1f427e8 fix(packaging): align MCP install paths with the patched SDK floor - e1cab5d0555824700e92a612ebda41b1f294eed3 fix(deps): require the patched MCP 1.30 transport - b21f3ecf6986e726c47d4db733b5378f70117c8f fix(spec): bound audit similarity and match policy words exactly - 1e90c47f6f97a4953d425b9cc41a54fb64a101c0 fix(import): exclude credentials from quick imports
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7dc9bb3add
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".
| setLayers(value) { | ||
| state.layers = value && typeof value === 'object' ? { ...value } : null; | ||
| explorerData = null; | ||
| uploadEdges(); |
There was a problem hiding this comment.
In the Every-node renderer, toggling a non-code relation layer reaches this method and immediately changes the rendered edges via uploadEdges(), but it never calls stats(). The Ledger handler does not refresh the count itself, so graphStatsChanged() retains the previous drawnLinks/hidden-relationship values until another unrelated action emits stats, leaving the graph summary inconsistent with the canvas; publish stats after the layer update as setScope() and setGhosts() do.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: faa65a4cac
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".
| flt, include_invalid=include_invalid, limit=500, prompt_only=prompt_only, | ||
| closed_only=closed_only, |
There was a problem hiding this comment.
When a workspace contains more than 500 closed memories, this call still caps list_memories at 500 records ordered by ingestion time before _relatedness scores them against the query. Although superseded filtering now happens server-side, fresh evidence is that an older exact-topic match is omitted when 501 newer unrelated closed records exist, causing “Only superseded versions” to report no history; rank or page the closed candidates before imposing this cap.
AGENTS.md reference: AGENTS.md:L175-L177
Useful? React with 👍 / 👎.
This branch has not been deployed
3 participants
Consolidates Classic capabilities into Ledger, redirects /classic, protects hidden drafts across refreshes and workspace changes, and enforces complete browser origins for token-free local APIs. Timeline filtering precedes limits and follows world/system time. This PR follows #261 and includes the verified #260/#261 fixes.
Community legend names and colors now remain stable when filtering changes their visible rank. Relation pagination focuses the first newly revealed row, including the final partial page. Existing draft, scope, temporal, CSP and origin review safeguards remain covered.
Validation: 95 focused graph, polling, spec API, evidence, MCP and skill tests; 13 documentation/distribution tests; the new relation-pagination Chromium regression; Ruff. Fresh immutable evidence, SVG/PNG provenance and README source links match this branch. Remote CI validates the complete matrix at the pushed head.