release: v0.9.3 — DX patch (feature-flag visibility + doctor command)#198
Merged
release: v0.9.3 — DX patch (feature-flag visibility + doctor command)#198
Conversation
Every disabled feature flag is now visible everywhere: viewer banner, CLI status, doctor output, and REST error responses. New `agentmemory doctor` command runs 10 diagnostic checks in one shot, each failure includes a concrete fix hint. Viewer gains first-run hero pointing at `demo` command and a footer with preset issue-report link. Bumps across 8 files per release checklist: package.json, plugin/.claude-plugin/plugin.json, src/version.ts, src/types.ts, src/functions/export-import.ts, test/export-import.test.ts, packages/mcp/package.json, CHANGELOG.md See CHANGELOG for full details.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Caution Review failedPull request was closed or merged during review 📝 WalkthroughWalkthroughAdds developer-experience features: CLI Changes
Sequence Diagram(s)sequenceDiagram
participant User as User/CLI
participant CLI as CLI Runner
participant Engine as Engine Service
participant Server as Server/Viewer
participant API as API (/agentmemory/config/flags)
participant Graph as Graph API
User->>CLI: agentmemory doctor
CLI->>Engine: GET /health
alt engine healthy
Engine-->>CLI: 200
else
CLI->>CLI: record failure / exit 1
end
CLI->>Server: GET /health
alt server healthy
Server-->>CLI: 200
else
CLI->>CLI: record failure / exit 1
end
CLI->>API: GET /agentmemory/config/flags
alt flags returned
API-->>CLI: flags + provider/embedding
CLI->>Graph: GET /agentmemory/graph/stats
alt graph populated & features enabled
Graph-->>CLI: nodeCount > 0
else
CLI->>CLI: record failure / exit 1
end
else
CLI->>CLI: record failure / exit 1
end
CLI->>User: consolidated pass/fail checklist
sequenceDiagram
participant Viewer as Browser Viewer
participant API as API Server
participant LocalStorage as localStorage
Viewer->>API: GET /agentmemory/config/flags
API-->>Viewer: flags array (affects, enableHow, docsHref, provider info)
Viewer->>Viewer: determine current tab
Viewer->>LocalStorage: read dismissed flags
loop per relevant flag
alt dismissed
Viewer-->>Viewer: skip rendering
else enabled
Viewer-->>Viewer: render info banner (collapsed by default)
else disabled
Viewer-->>Viewer: render collapsible warning with enableHow/docsHref
end
end
Viewer->>LocalStorage: persist dismiss/collapse state on user action
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
- Extract apiFetch() helper to deduplicate fetch/json/catch pattern - Drop redundant --port re-parse in getRestPort (env already set by outer block) - Unify safeDesc/safeCode in viewer banner render (single escHtml) - Extract formatChecks() + DoctorCheck type in runDoctor - Remove numbered step comments from runDoctor (flow speaks for itself) Zero behavior change.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 6
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/viewer/index.html (1)
889-916:⚠️ Potential issue | 🟠 MajorKeep the new banner/footer inside the scrollable layout.
#flag-bannersand#viewer-footerare siblings of the fixed-height.viewpanes whilebodystill hasoverflow: hidden. Because Line 171 still sizes each view for the old chrome, the added banner/footer push the bottom of the active tab below the viewport and make the footer unreachable on normal screen sizes.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/viewer/index.html` around lines 889 - 916, The flag banner (`#flag-banners`) and footer (`#viewer-footer`) are outside the scrollable area while .view panes are fixed-height and body has overflow:hidden, causing the active view content to be pushed below the viewport; fix by placing the banner and footer inside the scrollable layout or converting the page to a column flex layout where the header/banner, a central scrollable container for the .view panes (make that container flex:1 and overflow:auto), and the `#viewer-footer` are children so the .view content sizes correctly (use flex and/or calc heights rather than fixed heights) so the footer remains reachable and scrolls with view content.
🧹 Nitpick comments (4)
src/triggers/api.ts (3)
124-124: Minor: innerprovidershadows the outerregisterApiTriggersparameter.The
registerApiTriggerssignature on line 86 already has aprovider?: ResilientProvider | { circuitState?: unknown }parameter. Declaringconst provider = ...inside this handler shadows it. It's harmless here (the outerproviderisn't used in this callback), but a rename likeproviderKind/llmProviderKindwould prevent future foot-guns if someone later wants to surface the real circuit state in this response.♻️ Suggested rename
- const provider = env["ANTHROPIC_API_KEY"] || env["GEMINI_API_KEY"] || env["OPENROUTER_API_KEY"] || env["MINIMAX_API_KEY"] ? "llm" : "noop"; - const embeddingProvider = env["OPENAI_API_KEY"] || env["VOYAGE_API_KEY"] || env["COHERE_API_KEY"] || env["OLLAMA_HOST"] ? "embeddings" : "none"; + const providerKind = env["ANTHROPIC_API_KEY"] || env["GEMINI_API_KEY"] || env["OPENROUTER_API_KEY"] || env["MINIMAX_API_KEY"] ? "llm" : "noop"; + const embeddingProviderKind = env["OPENAI_API_KEY"] || env["VOYAGE_API_KEY"] || env["COHERE_API_KEY"] || env["OLLAMA_HOST"] ? "embeddings" : "none"; @@ - provider, - embeddingProvider, + provider: providerKind, + embeddingProvider: embeddingProviderKind,🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/triggers/api.ts` at line 124, The handler declares a local const provider that shadows the registerApiTriggers parameter named provider; rename the local variable (e.g., to providerKind or llmProviderKind) used in the expression `env["ANTHROPIC_API_KEY"] || ... ? "llm" : "noop"` and update any references to that local variable in the same callback so the outer parameter `provider` (from registerApiTriggers) is no longer shadowed.
1024-1030: Enriched error payloads — consider extracting to a helper.The four enriched 404 payloads (
api::graph-query,api::graph-stats,api::graph-extract,api::consolidate-pipeline) duplicate the same{ error, flag, enableHow, docsHref }shape, and the graph variants repeat the exact same four strings verbatim. Easy to drift over time (e.g., if the docs URL changes, it must be updated in three places).Consider a small factory — also makes it trivial to reuse when you later enrich the remaining unchanged
"not enabled"responses (e.g., Claude bridge, Snapshots, Team memory on lines 972, 993, 1147, 1167, 1185, 1271, 1290, 1311) so the structured-payload contract is uniform across the API.♻️ Suggested helper
function flagDisabledResponse(opts: { error: string; flag: string; enableHow: string; docsHref: string; }): Response { return { status_code: 404, body: { ...opts } }; } const graphDisabled = () => flagDisabledResponse({ error: "Knowledge graph not enabled", flag: "GRAPH_EXTRACTION_ENABLED", enableHow: "Set GRAPH_EXTRACTION_ENABLED=true and restart. Requires an LLM provider key.", docsHref: "https://github.com/rohitg00/agentmemory#knowledge-graph", });Also applies to: 1050-1056, 1085-1091, 1112-1118
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/triggers/api.ts` around lines 1024 - 1030, Extract a small factory like flagDisabledResponse(opts) that returns { status_code: 404, body: { ...opts } } and replace the repeated enriched 404 payload objects in the graph endpoints (the responses used by api::graph-query, api::graph-stats, api::graph-extract, and api::consolidate-pipeline) with calls to that factory (e.g., graphDisabled() that passes the four strings for error, flag, enableHow, docsHref); update the four duplicated literal objects to use this helper and reuse it for other similar "not enabled" responses noted in the comment to keep the structured payload consistent.
1024-1030: Nit:status_code: 404for a disabled-feature flag is semantically odd.A 404 implies "resource does not exist", but the route does exist — it's just gated off by config.
501 Not Implementedor503 Service Unavailable(with the same structured body) more accurately describes "this capability is turned off on this server", and the newflag/enableHow/docsHreffields already convey the remediation. The viewer/CLI parse the JSON body anyway, so changing the status wouldn't break them. Not blocking — raising it so the client-side banner logic isn't anchored to 404 forever.Also applies to: 1050-1056, 1085-1091, 1112-1118
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/triggers/api.ts` around lines 1024 - 1030, Replace the HTTP status_code for the "Knowledge graph not enabled" responses (the response objects that include flag: "GRAPH_EXTRACTION_ENABLED", enableHow, docsHref, etc.) from 404 to a more accurate status like 503 Service Unavailable (or 501 Not Implemented) to indicate the feature is disabled by configuration rather than the route missing; update all identical response objects (the other occurrences noted around the same blocks) so the client-side banner logic isn't tied to a 404.src/cli.ts (1)
494-559: Remove the numbered// 1.,// 2.comments inrunDoctor().They only restate what each block already does, so they're adding noise instead of structure here.
As per coding guidelines, "Avoid code comments explaining WHAT — use clear naming instead".
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/cli.ts` around lines 494 - 559, The numbered inline comments inside runDoctor() (e.g., "// 1. Server reachable", "// 2. Health status", "// 3. Viewer reachable", etc.) are redundant and should be removed; edit the runDoctor function to delete those numbered comment lines around the server/health/viewer/LLM/embedding/flags/graph checks so the code relies on clear variable and function names (serverUp, health, viewerUp, hasLlm, hasEmbed, checks) without the noisy "1./2./3." comments.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/cli.ts`:
- Around line 91-95: The doctor and status code are still hardcoding a localhost
viewer URL instead of using getBaseUrl(); update doctor() to probe the viewer
using getBaseUrl() (e.g., build the viewer probe URL from getBaseUrl() + the
viewer path the code expects) instead of `http://localhost:${port + 2}` and
update the status-printing logic (status() / printStatus()) to display the URL
returned by getBaseUrl() rather than a localhost-derived address; ensure you
reference getBaseUrl() as the single source of truth for the base viewer URL so
remote/reverse-proxied installs are handled correctly.
- Around line 507-560: runStatus() currently checks graph?.totalNodes but the
/agentmemory/graph/stats endpoint may return nodes or nodeCount; update the
graph population check to consider all returned fields (e.g., graph?.totalNodes
|| graph?.nodes || graph?.nodeCount) and coerce to a number before comparing > 0
so the "Knowledge graph populated" check (the graphHas calculation used when
pushing the "Knowledge graph populated" check) correctly reflects existing data.
In `@src/triggers/api.ts`:
- Around line 121-187: The config-flags endpoint (api::config-flags) is
currently unauthenticated and leaks sensitive metadata; fix by enforcing the
same auth as other endpoints: add the middleware "middleware::api-auth" to the
sdk.registerTrigger config and/or call checkAuth(req,
process.env.AGENTMEMORY_SECRET) at the top of the api::config-flags handler
(before computing VERSION/provider/flags) so requests without a valid bearer
secret are rejected; reference the existing checkAuth helper, the
AGENTMEMORY_SECRET env var, and sdk.registerTrigger to mirror the
health/sessions endpoints' pattern.
- Around line 124-125: The embeddingProvider detection in the status logic
incorrectly omits OPENROUTER_API_KEY; update the expression that assigns
embeddingProvider (the variable in src/triggers/api.ts) to include
env["OPENROUTER_API_KEY"] alongside OPENAI_API_KEY, VOYAGE_API_KEY,
COHERE_API_KEY, and OLLAMA_HOST so it returns "embeddings" when only OpenRouter
is configured; ensure this matches the behavior of detectEmbeddingProvider() in
src/config.ts.
In `@src/viewer/index.html`:
- Around line 591-599: The clickable .flag-summary currently uses a div which is
not keyboard-focusable—replace the interactive div with a real <button
class="flag-summary" type="button"> (and do the same at the other occurrence
referenced) so it is natively focusable and operable by keyboard; keep the
existing .flag-summary CSS (add appearance: none; border: none; background: none
if needed to preserve visual styles), update any JS that reads/writes the
expanded state to use the button's aria-expanded attribute (e.g., toggle
aria-expanded="true"/"false") and ensure existing click handlers remain attached
to the button element.
- Around line 589-650: The new banner/card styles (.flag-summary, .flag-banner)
use var(--bg-subtle) which is not defined; add a --bg-subtle CSS custom property
to both theme blocks (light and dark) where other tokens live (e.g., alongside
--bg, --bg-alt, --border) so the background declarations don't fall back to
transparent; choose a subtle background color consistent with each theme (light:
a slightly lighter tint than --bg, dark: a slightly darker tint than --bg) and
place the definitions in the same theme selector so .flag-summary and
.flag-banner inherit them.
---
Outside diff comments:
In `@src/viewer/index.html`:
- Around line 889-916: The flag banner (`#flag-banners`) and footer
(`#viewer-footer`) are outside the scrollable area while .view panes are
fixed-height and body has overflow:hidden, causing the active view content to be
pushed below the viewport; fix by placing the banner and footer inside the
scrollable layout or converting the page to a column flex layout where the
header/banner, a central scrollable container for the .view panes (make that
container flex:1 and overflow:auto), and the `#viewer-footer` are children so the
.view content sizes correctly (use flex and/or calc heights rather than fixed
heights) so the footer remains reachable and scrolls with view content.
---
Nitpick comments:
In `@src/cli.ts`:
- Around line 494-559: The numbered inline comments inside runDoctor() (e.g.,
"// 1. Server reachable", "// 2. Health status", "// 3. Viewer reachable", etc.)
are redundant and should be removed; edit the runDoctor function to delete those
numbered comment lines around the server/health/viewer/LLM/embedding/flags/graph
checks so the code relies on clear variable and function names (serverUp,
health, viewerUp, hasLlm, hasEmbed, checks) without the noisy "1./2./3."
comments.
In `@src/triggers/api.ts`:
- Line 124: The handler declares a local const provider that shadows the
registerApiTriggers parameter named provider; rename the local variable (e.g.,
to providerKind or llmProviderKind) used in the expression
`env["ANTHROPIC_API_KEY"] || ... ? "llm" : "noop"` and update any references to
that local variable in the same callback so the outer parameter `provider` (from
registerApiTriggers) is no longer shadowed.
- Around line 1024-1030: Extract a small factory like flagDisabledResponse(opts)
that returns { status_code: 404, body: { ...opts } } and replace the repeated
enriched 404 payload objects in the graph endpoints (the responses used by
api::graph-query, api::graph-stats, api::graph-extract, and
api::consolidate-pipeline) with calls to that factory (e.g., graphDisabled()
that passes the four strings for error, flag, enableHow, docsHref); update the
four duplicated literal objects to use this helper and reuse it for other
similar "not enabled" responses noted in the comment to keep the structured
payload consistent.
- Around line 1024-1030: Replace the HTTP status_code for the "Knowledge graph
not enabled" responses (the response objects that include flag:
"GRAPH_EXTRACTION_ENABLED", enableHow, docsHref, etc.) from 404 to a more
accurate status like 503 Service Unavailable (or 501 Not Implemented) to
indicate the feature is disabled by configuration rather than the route missing;
update all identical response objects (the other occurrences noted around the
same blocks) so the client-side banner logic isn't tied to a 404.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 2a63fe94-2d7f-4149-b79e-e9ec8beb9bf5
📒 Files selected for processing (13)
CHANGELOG.mdpackage.jsonpackages/mcp/package.jsonplugin/.claude-plugin/plugin.jsonsrc/cli.tssrc/functions/export-import.tssrc/triggers/api.tssrc/types.tssrc/version.tssrc/viewer/index.htmltest/export-import.test.tswebsite/components/Install.tsxwebsite/lib/generated-meta.json
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
src/cli.ts (1)
524-530:⚠️ Potential issue | 🟠 Major
runDoctorregressions still present from earlier review.Two issues from prior review rounds are still in this revision:
- Line 524 — Viewer probe is hardcoded to
http://localhost:${viewerPort}. For any non-localAGENTMEMORY_URL(or a URL without an explicit port —getRestPort()falls back to3111becausenew URL("https://x").port === ""), this check will always fail andviewerPortitself is meaningless. Reuse a viewer URL derived fromgetBaseUrl().- Line 530 —
graphHasonly inspectsgraph?.totalNodes, whilerunStatus(line 438) and the viewer readgraph?.nodes/graph?.nodeCount. Doctor will report "Knowledge graph populated" as failing on healthy installs that have data.🛠️ Suggested fix
- const viewerUp = await fetch(`http://localhost:${viewerPort}`, { signal: AbortSignal.timeout(2000) }) + const viewerUrl = new URL(`:${viewerPort}`, base).toString().replace(/\/$/, ""); + const viewerUp = await fetch(viewerUrl, { signal: AbortSignal.timeout(2000) }) .then((r) => r.ok) .catch(() => false); @@ - const graphHas = (graph?.totalNodes || 0) > 0; + const graphNodeCount = graph?.nodes ?? graph?.nodeCount ?? graph?.totalNodes ?? 0; + const graphHas = graphNodeCount > 0;Also consider plumbing
viewerUrlinto the failure hint on line 541 instead of justPort ${viewerPort} not responding.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/cli.ts` around lines 524 - 530, The viewer probe incorrectly uses a hardcoded localhost URL and viewerPort; change the probe to use the actual viewer base URL from getBaseUrl() (or construct viewerUrl via getBaseUrl/getRestPort logic) instead of `http://localhost:${viewerPort}`, and update the failure hint to reference viewerUrl rather than just the port (symbols: viewerPort, getBaseUrl, getRestPort, viewerUrl). Also fix graph detection: replace the single check `graphHas = (graph?.totalNodes || 0) > 0` with a more complete check that considers graph?.nodes and graph?.nodeCount (the same criteria used by runStatus and the viewer) so `graphHas` accurately reflects a populated knowledge graph (symbols: graphHas, graph?.nodes, graph?.nodeCount, runStatus). Ensure these changes are applied where runDoctor constructs viewer probe and graphHas.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/cli.ts`:
- Around line 404-411: The apiFetch function currently returns res.json() for
all HTTP responses which treats 4xx/5xx error payloads as successful results;
update apiFetch (the async function apiFetch<T = unknown>(base: string, path:
string, timeoutMs = 5000)) to check res.ok before returning the parsed body—if
res.ok is false, parse the JSON error body (or text) to capture details and then
return null (or propagate a structured error) so callers like
runStatus/runDoctor don't receive malformed success objects; ensure the logic
uses the same AbortSignal.timeout and still handles network exceptions in the
existing catch path.
---
Duplicate comments:
In `@src/cli.ts`:
- Around line 524-530: The viewer probe incorrectly uses a hardcoded localhost
URL and viewerPort; change the probe to use the actual viewer base URL from
getBaseUrl() (or construct viewerUrl via getBaseUrl/getRestPort logic) instead
of `http://localhost:${viewerPort}`, and update the failure hint to reference
viewerUrl rather than just the port (symbols: viewerPort, getBaseUrl,
getRestPort, viewerUrl). Also fix graph detection: replace the single check
`graphHas = (graph?.totalNodes || 0) > 0` with a more complete check that
considers graph?.nodes and graph?.nodeCount (the same criteria used by runStatus
and the viewer) so `graphHas` accurately reflects a populated knowledge graph
(symbols: graphHas, graph?.nodes, graph?.nodeCount, runStatus). Ensure these
changes are applied where runDoctor constructs viewer probe and graphHas.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: b83a983d-d10c-4759-bca3-d40f44089566
📒 Files selected for processing (2)
src/cli.tssrc/viewer/index.html
✅ Files skipped from review due to trivial changes (1)
- src/viewer/index.html
Comment on lines
+404
to
+411
| async function apiFetch<T = unknown>(base: string, path: string, timeoutMs = 5000): Promise<T | null> { | ||
| try { | ||
| const res = await fetch(`${base}/agentmemory/${path}`, { signal: AbortSignal.timeout(timeoutMs) }); | ||
| return (await res.json()) as T; | ||
| } catch { | ||
| return null; | ||
| } | ||
| } |
There was a problem hiding this comment.
apiFetch ignores HTTP error responses.
apiFetch calls res.json() without first checking res.ok, so a 4xx/5xx response with a JSON error body (e.g., the new structured { error, flag, enableHow, docsHref } payload from "not enabled" routes) is returned as a successful result. This silently feeds malformed objects into runStatus/runDoctor (e.g., flags?.flags becomes undefined, health?.status is missing, doctor reports green/red incorrectly).
🛡️ Suggested fix
async function apiFetch<T = unknown>(base: string, path: string, timeoutMs = 5000): Promise<T | null> {
try {
const res = await fetch(`${base}/agentmemory/${path}`, { signal: AbortSignal.timeout(timeoutMs) });
+ if (!res.ok) return null;
return (await res.json()) as T;
} catch {
return null;
}
}🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@src/cli.ts` around lines 404 - 411, The apiFetch function currently returns
res.json() for all HTTP responses which treats 4xx/5xx error payloads as
successful results; update apiFetch (the async function apiFetch<T =
unknown>(base: string, path: string, timeoutMs = 5000)) to check res.ok before
returning the parsed body—if res.ok is false, parse the JSON error body (or
text) to capture details and then return null (or propagate a structured error)
so callers like runStatus/runDoctor don't receive malformed success objects;
ensure the logic uses the same AbortSignal.timeout and still handles network
exceptions in the existing catch path.
- cli: use getBaseUrl()-derived viewer URL instead of hardcoded localhost (new getViewerUrl() helper; honors AGENTMEMORY_VIEWER_URL + base hostname) - cli: null-safe graph count (totalNodes || nodes || nodeCount) in status + doctor — matches viewer dashboard fallback - api: auth config-flags via middleware::api-auth + explicit checkAuth (leaked provider + version state previously) - api: use detectEmbeddingProvider() from config instead of hand-rolled key check (was missing GEMINI_API_KEY + OPENROUTER_API_KEY; now matches config.ts single-source-of-truth) - api: rename local 'provider' to 'providerKind' to avoid shadowing registerApiTriggers() parameter - api: extract flagDisabledResponse() factory + graphDisabledResponse() + consolidationDisabledResponse() helpers; status_code 404 -> 503 (disabled-by-config != missing route) - viewer: add --bg-subtle token to both light + dark themes (6 call sites previously fell back to transparent) - viewer: .flag-summary div -> button with aria-expanded + aria-controls for keyboard a11y - viewer: body flex column + .view flex:1 min-height:0 so banner + footer sit outside scroll region without clipping view content - viewer: dashboard + graph tab prefer totalNodes/totalEdges (matches actual mem::graph-stats response; pre-existing fallback bug) Zero behavior change in happy path. Fixes latent correctness + a11y issues.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Developer-experience patch. Every disabled feature flag is now visible in the viewer, the CLI, and REST error responses, so devs no longer hit empty tabs wondering whether the install is broken or just opt-in. Adds a
doctorcommand that diagnoses the whole stack in one shot and a first-run hero in the viewer that points at the magical-momentdemocommand.Changes
New surfaces
agentmemory doctor— 10 diagnostic checks with fix hints on every failureGET /agentmemory/config/flags— structured flag/provider/embedding statedemocommand prompt when sessions emptyagentmemory viewer · v{ver} · github · docs · report issue →with preset GitHub issue bodyEnriched
statuscommand shows provider + embedding + per-flag tick/crossAGENTMEMORY_URLenv now honored (was documented but ignored){ error, flag, enableHow, docsHref }matching viewer banner contractWebsite
demoto step 2 (magical moment on the critical path)gen-meta.mjsprebuild (no more stale version drift)Version bumps across 8 files per release checklist:
package.json,plugin/.claude-plugin/plugin.json,src/version.tssrc/types.ts(ExportData union),src/functions/export-import.ts(supportedVersions)test/export-import.test.ts,packages/mcp/package.json,CHANGELOG.mdTest plan
npm run build— clean buildnpm test -- test/consistency.test.ts test/export-import.test.ts— 13/13 passagentmemory doctor— 10 checks render correctly, hints shown on failuresagentmemory status— provider + embedding + flag table rendersAGENTMEMORY_URL=http://localhost:3111 agentmemory status— env honoredcurl /agentmemory/graph/statswith flag OFF — returns{ error, flag, enableHow, docsHref }CHANGELOG
See
CHANGELOG.md[0.9.3] section for the complete list.Summary by CodeRabbit
agentmemory doctordiagnostics, new REST endpoint for feature-flag metadata, and viewer banners with collapsible, dismissible, per-flag persistence and tab filtering.democommand.