Conversation
…es for API spec v0.6.7 - New Deep Search API with sync, streaming (SSE), and background modes - Data connector file browsing endpoint with provider-specific filtering - Describe job update endpoint (PATCH) for use_in_default_index flag - Response API: flexible knowledge base (collections/files/default), tools, max_output_tokens - Updated generate.js to handle nullable Partial types and discriminatedUnion fixes Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
📝 WalkthroughWalkthroughAdds Deep Search streaming and non‑streaming APIs, data‑connector file listing, Response API knowledge‑base/tool types, client wiring for Deep Search, and extensive post‑generation TypeScript/Zod postprocessing including File import aliasing and nullability fixes. Changes
Sequence DiagramsequenceDiagram
participant Client
participant EnhancedAPI as EnhancedDeepSearchApi
participant Fetch as fetch/SSE
participant Backend as BackendService
participant Consumer as EventConsumer
Client->>EnhancedAPI: createStreamingDeepSearch(params)
activate EnhancedAPI
EnhancedAPI->>Fetch: POST /api/v1/deep-searches (stream=true)
activate Fetch
Fetch->>Backend: HTTP request with SSE headers
activate Backend
Backend-->>Fetch: ReadableStream<Uint8Array> (SSE)
deactivate Backend
Fetch-->>EnhancedAPI: Stream opened
deactivate Fetch
EnhancedAPI->>EnhancedAPI: parseSSEStream(stream)
loop stream processing
EnhancedAPI->>EnhancedAPI: processLines(buffer)
EnhancedAPI->>Consumer: yield events (Created / TextDelta / ResultAdded / Completed / Error)
end
EnhancedAPI->>Consumer: final event / close
deactivate EnhancedAPI
Client-->>Consumer: consumes AsyncIterable events
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
📝 Coding Plan
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 |
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
package.json (1)
36-36:⚠️ Potential issue | 🔴 CriticalUpgrade axios to ^1.13.5 to address CVE-2026-25639.
axios 1.12.0 is affected by CVE-2026-25639, a DoS vulnerability in the
mergeConfigfunction when processing objects with__proto__. Upgrade to version 1.13.5 or higher to patch this issue.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@package.json` at line 36, Update the axios dependency in package.json from "axios": "^1.12.0" to "axios": "^1.13.5" to remediate CVE-2026-25639; after changing the version string, run your package manager to update node_modules and the lockfile (npm install / pnpm install / yarn install), regenerate the lockfile commit, and run tests and a dependency audit (npm audit / pnpm audit) to verify no regressions.
🧹 Nitpick comments (2)
src/api/deep-search.api.ts (1)
283-290: Non-null assertion onstatusmay mask issues if the API response is malformed.The code uses
deepSearch.status!with a non-null assertion. If the API returns a response without astatusfield (e.g., due to a schema mismatch or unexpected error), this could lead to unexpected behavior where the polling loop continues indefinitely.Consider adding a defensive check:
🛡️ Suggested defensive check
const deepSearch = await this.getDeepSearch(deepSearchId); + + if (!deepSearch.status) { + throw new CloudglueError( + `Deep search ${deepSearchId} returned invalid response without status`, + ); + } - if (['completed', 'failed', 'cancelled'].includes(deepSearch.status!)) { + if (['completed', 'failed', 'cancelled'].includes(deepSearch.status)) {🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/api/deep-search.api.ts` around lines 283 - 290, The code uses a non-null assertion on deepSearch.status which can hide missing/undefined statuses and cause the poll to hang; update the logic in the function that handles the poll response (referencing deepSearch.status, deepSearchId, and CloudglueError) to explicitly check whether deepSearch.status is null/undefined before using it, and if it is missing throw a CloudglueError (including deepSearchId and any available response/error info) so the polling loop fails fast instead of continuing indefinitely; remove the `!` assertion and replace it with a defensive null/undefined guard that either throws or returns an appropriate error result.package.json (1)
3-3: Consider a minor version bump instead of a patch for new features.The PR adds new API surfaces (Deep Search API, data connector file browsing) and new functionality (describe updates, response API enhancements). According to semantic versioning, new features typically warrant a minor version bump (e.g., 0.8.0) rather than a patch bump (0.7.4), which is typically reserved for backward-compatible bug fixes.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@package.json` at line 3, The package.json currently sets "version": "0.7.4" but new features were added; update the version field to a minor bump (e.g., "0.8.0") to reflect added APIs (Deep Search, data connector browsing, response API enhancements) and ensure any release notes/changelog entries reflect the new minor release; update package.json's "version" value and verify any CI/release scripts that read the version are in sync.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@package.json`:
- Line 3: The package.json "version" field currently reads "0.7.4" but the PR
title references v0.6.7; reconcile this by deciding the correct release version
and updating the mismatch: either change the package.json "version" value to
"0.6.7" or update the PR title to v0.7.4 so they match, and ensure any
accompanying changelog/release notes reflect the chosen version.
In `@spec`:
- Line 1: The PR references a submodule commit for the spec that doesn't exist,
breaking generate.js which relies on spec/spec/openapi.json; fix by either (A)
creating and pushing the missing commit in the cloudglue-api-spec repository so
its object ID matches b522b9959ddf32bedeba525d7afd0cbb7e517169 and ensure
spec/spec/openapi.json is present in that commit, or (B) update the spec
submodule pointer in this repo to a valid, public commit in cloudglue-api-spec
that contains spec/spec/openapi.json and commit the updated submodule reference
(and update .gitmodules if needed) so generate.js can run. Ensure the chosen
commit is publicly accessible before merging.
---
Outside diff comments:
In `@package.json`:
- Line 36: Update the axios dependency in package.json from "axios": "^1.12.0"
to "axios": "^1.13.5" to remediate CVE-2026-25639; after changing the version
string, run your package manager to update node_modules and the lockfile (npm
install / pnpm install / yarn install), regenerate the lockfile commit, and run
tests and a dependency audit (npm audit / pnpm audit) to verify no regressions.
---
Nitpick comments:
In `@package.json`:
- Line 3: The package.json currently sets "version": "0.7.4" but new features
were added; update the version field to a minor bump (e.g., "0.8.0") to reflect
added APIs (Deep Search, data connector browsing, response API enhancements) and
ensure any release notes/changelog entries reflect the new minor release; update
package.json's "version" value and verify any CI/release scripts that read the
version are in sync.
In `@src/api/deep-search.api.ts`:
- Around line 283-290: The code uses a non-null assertion on deepSearch.status
which can hide missing/undefined statuses and cause the poll to hang; update the
logic in the function that handles the poll response (referencing
deepSearch.status, deepSearchId, and CloudglueError) to explicitly check whether
deepSearch.status is null/undefined before using it, and if it is missing throw
a CloudglueError (including deepSearchId and any available response/error info)
so the polling loop fails fast instead of continuing indefinitely; remove the
`!` assertion and replace it with a defensive null/undefined guard that either
throws or returns an appropriate error result.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: d89fe696-d042-4c94-8b45-0b38f15c397e
⛔ Files ignored due to path filters (7)
generated/Data_Connectors.tsis excluded by!**/generated/**generated/Deep_Search.tsis excluded by!**/generated/**generated/Describe.tsis excluded by!**/generated/**generated/Response.tsis excluded by!**/generated/**generated/common.tsis excluded by!**/generated/**generated/index.tsis excluded by!**/generated/**package-lock.jsonis excluded by!**/package-lock.json
📒 Files selected for processing (9)
generate.jspackage.jsonspecsrc/api/data-connectors.api.tssrc/api/deep-search.api.tssrc/api/describe.api.tssrc/api/response.api.tssrc/client.tssrc/types.ts
| { | ||
| "name": "@cloudglue/cloudglue-js", | ||
| "version": "0.7.3", | ||
| "version": "0.7.4", |
There was a problem hiding this comment.
Version mismatch: PR title references v0.6.7 but package.json shows 0.7.4.
The PR title mentions "v0.6.7" but the version field shows "0.7.4". Please verify which version number is correct and update either the PR title or the package.json to ensure consistency.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@package.json` at line 3, The package.json "version" field currently reads
"0.7.4" but the PR title references v0.6.7; reconcile this by deciding the
correct release version and updating the mismatch: either change the
package.json "version" value to "0.6.7" or update the PR title to v0.7.4 so they
match, and ensure any accompanying changelog/release notes reflect the chosen
version.
| @@ -1 +1 @@ | |||
| Subproject commit 1f9df1729b97053112d2caac753545f5e4858e75 | |||
| Subproject commit b522b9959ddf32bedeba525d7afd0cbb7e517169 | |||
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Description: Verify the spec submodule commit exists and show the changes
# Check if submodule directory exists and show commit details
if [ -d "spec" ]; then
cd spec
# Verify the new commit exists
echo "=== Verifying new commit exists ==="
git cat-file -t b522b9959ddf32bedeba525d7afd0cbb7e517169 2>/dev/null && echo "✓ Commit exists" || echo "✗ Commit not found"
# Show commit message
echo -e "\n=== New commit details ==="
git log -1 --oneline b522b9959ddf32bedeba525d7afd0cbb7e517169
# Show what changed between old and new commit (if old commit is known)
echo -e "\n=== Files changed in submodule ==="
git diff --name-status 1f9df1729b97053112d2caac753545f5e4858e75..b522b9959ddf32bedeba525d7afd0cbb7e517169
cd ..
else
echo "Submodule directory 'spec' not found"
fiRepository: cloudglue/cloudglue-js
Length of output: 545
🏁 Script executed:
#!/bin/bash
# Description: Examine submodule configuration and status
echo "=== Checking .gitmodules file ==="
cat .gitmodules 2>/dev/null || echo "No .gitmodules file found"
echo -e "\n=== Checking git submodule status ==="
git config --file=.gitmodules --get-all submodule.spec.url 2>/dev/null || echo "No spec submodule found in config"
echo -e "\n=== Checking spec directory ==="
ls -la spec 2>/dev/null | head -20
echo -e "\n=== Checking git index for spec submodule ==="
git ls-files --stage spec | head -10
echo -e "\n=== Checking if spec is a gitlink ==="
git cat-file -p HEAD:spec 2>/dev/null || echo "spec not found in HEAD"
echo -e "\n=== Checking git log for spec file/module ==="
git log --oneline -- spec 2>/dev/null | head -5Repository: cloudglue/cloudglue-js
Length of output: 726
🏁 Script executed:
#!/bin/bash
# Description: Check recent commits and understand the submodule update context
echo "=== Recent commits in cloudglue-js ==="
git log --oneline -10
echo -e "\n=== Commit that updated spec submodule (if recent) ==="
git log --oneline --all -- spec | head -5
echo -e "\n=== Check if new commit b522b99... is in any branch ==="
git branch -r --contains b522b9959ddf32bedeba525d7afd0cbb7e517169 2>/dev/null | head -5 || echo "Commit not found in any remote branch"
echo -e "\n=== Show the spec submodule change in current commit ==="
git diff HEAD^ HEAD -- spec 2>/dev/null | head -20
echo -e "\n=== Check .gitmodules changes ==="
git log --oneline -p -- .gitmodules | head -50Repository: cloudglue/cloudglue-js
Length of output: 1120
🏁 Script executed:
#!/bin/bash
# Description: Get more context about the spec submodule update
echo "=== Current working directory and HEAD ==="
pwd
git rev-parse HEAD
git describe --always
echo -e "\n=== Check if spec directory has any git history ==="
cd spec && git log --oneline 2>&1 | head -10
cd ..
echo -e "\n=== List all files that were changed in the PR commit ==="
git show --name-only 54d3f13 | head -30
echo -e "\n=== Check if there are generated files that might depend on spec ==="
find . -name "*.ts" -o -name "*.js" | head -20 | xargs ls -la 2>/dev/null | head -30Repository: cloudglue/cloudglue-js
Length of output: 2901
🏁 Script executed:
#!/bin/bash
# Description: Check generate.js to understand spec dependency
echo "=== Head of generate.js to understand spec path ==="
head -50 generate.js
echo -e "\n=== Search for 'spec' references in generate.js ==="
grep -n "spec" generate.js | head -20
echo -e "\n=== Check package.json scripts ==="
cat package.json | grep -A 5 "\"scripts\""
echo -e "\n=== Check if spec is in .gitignore ==="
grep -i "spec" .gitignoreRepository: cloudglue/cloudglue-js
Length of output: 3338
The submodule commit does not exist in the cloudglue-api-spec repository.
The spec submodule is configured to reference commit b522b9959ddf32bedeba525d7afd0cbb7e517169, but this commit does not exist in git@github.com:cloudglue/cloudglue-api-spec.git. The PR's generate.js script depends on spec/spec/openapi.json from this submodule to generate the API client code. Without the submodule commit being available, the project cannot be cloned and built, and the generated code cannot be regenerated or verified.
Ensure the corresponding commit in the cloudglue-api-spec repository is created, pushed, and publicly accessible before this PR can be merged.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@spec` at line 1, The PR references a submodule commit for the spec that
doesn't exist, breaking generate.js which relies on spec/spec/openapi.json; fix
by either (A) creating and pushing the missing commit in the cloudglue-api-spec
repository so its object ID matches b522b9959ddf32bedeba525d7afd0cbb7e517169 and
ensure spec/spec/openapi.json is present in that commit, or (B) update the spec
submodule pointer in this repo to a valid, public commit in cloudglue-api-spec
that contains spec/spec/openapi.json and commit the updated submodule reference
(and update .gitmodules if needed) so generate.js can run. Ensure the chosen
commit is publicly accessible before merging.
amyxst
approved these changes
Mar 16, 2026
…lity Ensures unknown SSE event types from the server aren't silently dropped. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 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/api/deep-search.api.ts`:
- Around line 189-199: The fetch call builds headers from
this.api.axios.defaults.headers and directly uses h['Authorization'],
h['x-sdk-client'], h['x-sdk-version'], which may be undefined; change the header
construction used in the fetch call to build a headers object dynamically (e.g.,
start with required static headers 'Content-Type' and 'Accept' then add
Authorization, x-sdk-client, x-sdk-version only if truthy) so undefined values
are not passed to fetch; locate the header read in the method that calls fetch
(the lines using const h = this.api.axios.defaults.headers and the subsequent
fetch(... headers: { ... })) and update it to conditionally include or filter
out undefined header entries.
- Around line 286-296: The loop that polls getDeepSearch uses a non-null
assertion on deepSearch.status and lets undefined fall through, which can cause
infinite polling; update the logic in the polling loop (where getDeepSearch is
called and deepSearch.status is inspected) to explicitly check for the presence
of deepSearch.status and fail fast if it's missing (for example: if status is
undefined/null throw a CloudglueError referencing deepSearchId and any
deepSearch.error), remove the non-null assertion usage, and ensure the existing
failure branch (deepSearch.status === 'failed') and the completed/cancelled
branch still work as intended.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 46c6a4d3-d7dc-4489-bd25-8b60dcee8508
📒 Files selected for processing (1)
src/api/deep-search.api.ts
…p search - Add null check on deepSearch.status in waitForReady to fail fast instead of polling indefinitely if the API returns a malformed response - Build fetch headers dynamically to avoid passing undefined values Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
src/api/deep-search.api.ts (1)
42-88:DeepSearchStreamEventTypeis not actually discriminated.Including
DeepSearchGenericEvent { type: string }in the public union prevents useful narrowing onevent.type, and theRecord<string, any>payloads throw away the concreteDeepSearch/DeepSearchResultshapes. If you need an unknown-event fallback, normalize it to a separate'unknown'variant instead of unioning intype: string.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/api/deep-search.api.ts` around lines 42 - 88, The union is not discriminated because DeepSearchGenericEvent uses type: string and the payloads use broad Record<string, any>; update the specific event interfaces (DeepSearchCreatedEvent, DeepSearchResultAddedEvent, DeepSearchCompletedEvent) to reference concrete types (e.g., DeepSearch and DeepSearchResult) instead of Record<string, any>, and change DeepSearchGenericEvent to a single fallback with type: 'unknown' and a generic payload field (e.g., payload: Record<string, any>) so narrowing on event.type works; then keep DeepSearchStreamEventType as the union of those concrete event interfaces plus the new DeepSearchGenericEvent('unknown') fallback.
🤖 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/api/deep-search.api.ts`:
- Around line 13-30: The exported CreateDeepSearchParams type currently includes
the stream?: boolean property which allows callers of createDeepSearch() (and
client.deepSearch.createDeepSearch) to compile with streaming enabled even
though only createStreamingDeepSearch() can handle SSE; remove the stream
property from CreateDeepSearchParams and any other non-streaming request types
(also update the duplicate definitions referenced around the other occurrence)
so that createDeepSearch()’s signature cannot accept stream: true, leaving
streaming behavior only to createStreamingDeepSearch() and its corresponding
request type.
- Around line 186-223: The fetch call and subsequent stream consumption in the
deep-search endpoint can throw transport-level errors that currently escape;
wrap the entire network interaction (the fetch(...) call and any downstream
stream-reading logic) in a try/catch and, on any thrown error, rethrow a
CloudglueError so transport failures are normalized with HTTP errors.
Specifically, catch errors around the fetch invocation and stream processing in
the method containing the fetch, build a CloudglueError using a sensible status
(e.g., 0 or undefined), include JSON.stringify(body) as the request payload,
include headers (or an empty object) from response.headers if available, and
attach the original error object/string as the error payload/responseData;
ensure you still preserve existing behavior for non-ok HTTP responses handled by
the existing response.ok branch.
---
Nitpick comments:
In `@src/api/deep-search.api.ts`:
- Around line 42-88: The union is not discriminated because
DeepSearchGenericEvent uses type: string and the payloads use broad
Record<string, any>; update the specific event interfaces
(DeepSearchCreatedEvent, DeepSearchResultAddedEvent, DeepSearchCompletedEvent)
to reference concrete types (e.g., DeepSearch and DeepSearchResult) instead of
Record<string, any>, and change DeepSearchGenericEvent to a single fallback with
type: 'unknown' and a generic payload field (e.g., payload: Record<string, any>)
so narrowing on event.type works; then keep DeepSearchStreamEventType as the
union of those concrete event interfaces plus the new
DeepSearchGenericEvent('unknown') fallback.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 966400e0-cbb2-486a-9382-43f2b7ad8218
📒 Files selected for processing (1)
src/api/deep-search.api.ts
Comment on lines
+13
to
+30
| export interface CreateDeepSearchParams { | ||
| /** Knowledge base configuration specifying what content to search */ | ||
| knowledge_base: DeepSearchKnowledgeBase; | ||
| /** The search query */ | ||
| query: string; | ||
| /** Search scope - segment-level or file-level */ | ||
| scope?: 'segment' | 'file'; | ||
| /** Maximum number of results to return (1-500) */ | ||
| limit?: number; | ||
| /** Whether to exclude weak results */ | ||
| exclude_weak_results?: boolean; | ||
| /** Include additional data in results */ | ||
| include?: Array<'search_queries'>; | ||
| /** Enable server-sent events streaming */ | ||
| stream?: boolean; | ||
| /** Run the deep search in background (async) */ | ||
| background?: boolean; | ||
| } |
There was a problem hiding this comment.
Remove stream from the createDeepSearch() contract.
createDeepSearch() forwards to the generated JSON client, so client.deepSearch.createDeepSearch({ ..., stream: true }) compiles even though only createStreamingDeepSearch() can consume an SSE response. Tighten the exported params/signature so the non-streaming entry point cannot be called in streaming mode.
Also applies to: 166-168
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@src/api/deep-search.api.ts` around lines 13 - 30, The exported
CreateDeepSearchParams type currently includes the stream?: boolean property
which allows callers of createDeepSearch() (and
client.deepSearch.createDeepSearch) to compile with streaming enabled even
though only createStreamingDeepSearch() can handle SSE; remove the stream
property from CreateDeepSearchParams and any other non-streaming request types
(also update the duplicate definitions referenced around the other occurrence)
so that createDeepSearch()’s signature cannot accept stream: true, leaving
streaming behavior only to createStreamingDeepSearch() and its corresponding
request type.
Comment on lines
+186
to
+223
| const baseURL = this.api.axios.defaults.baseURL; | ||
| const url = `${baseURL}/deepSearch`; | ||
|
|
||
| const h = this.api.axios.defaults.headers as Record<string, any>; | ||
| const headers: Record<string, string> = { | ||
| 'Content-Type': 'application/json', | ||
| Accept: 'text/event-stream', | ||
| }; | ||
| if (h['Authorization']) headers['Authorization'] = h['Authorization']; | ||
| if (h['x-sdk-client']) headers['x-sdk-client'] = h['x-sdk-client']; | ||
| if (h['x-sdk-version']) headers['x-sdk-version'] = h['x-sdk-version']; | ||
|
|
||
| const response = await fetch(url, { | ||
| method: 'POST', | ||
| headers, | ||
| body: JSON.stringify(body), | ||
| }); | ||
|
|
||
| if (!response.ok) { | ||
| let errorMessage = `Request failed with status ${response.status}`; | ||
| let responseData: any; | ||
| try { | ||
| responseData = await response.json(); | ||
| if (typeof responseData?.error === 'string') { | ||
| errorMessage = responseData.error; | ||
| } else if (responseData?.error?.message) { | ||
| errorMessage = responseData.error.message; | ||
| } | ||
| } catch { | ||
| // Could not parse error body | ||
| } | ||
| throw new CloudglueError( | ||
| errorMessage, | ||
| response.status, | ||
| JSON.stringify(body), | ||
| Object.fromEntries(response.headers.entries()), | ||
| responseData, | ||
| ); |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's examine the full context of the createStreamingDeepSearch method
head -230 src/api/deep-search.api.ts | tail -60Repository: cloudglue/cloudglue-js
Length of output: 2097
🏁 Script executed:
# Check the full file structure
wc -l src/api/deep-search.api.tsRepository: cloudglue/cloudglue-js
Length of output: 96
🏁 Script executed:
# Search for streaming patterns in the codebase
rg -i "streaming|event-stream|sse" --type ts src/ -A 3 -B 1Repository: cloudglue/cloudglue-js
Length of output: 10083
🏁 Script executed:
# Check if there are other fetch() calls in src/api
rg "fetch\(" src/api --type ts -B 2 -A 2Repository: cloudglue/cloudglue-js
Length of output: 715
🏁 Script executed:
# Check the Zodios usage in the codebase to understand how it's configured
rg "zodios" src/api --type ts -B 2 -A 2 | head -40Repository: cloudglue/cloudglue-js
Length of output: 48
🏁 Script executed:
# Look for any comments or documentation about streaming and transport choices
rg -i "zodios.*stream|stream.*zodios|sse.*zodios" src/ --type ts -B 2 -A 2Repository: cloudglue/cloudglue-js
Length of output: 48
🏁 Script executed:
# Check if there's any configuration or setup guidance in README or similar
fd -t f "README|CLAUDE|\.md" . -exec grep -l "stream\|sse\|fetch" {} \;Repository: cloudglue/cloudglue-js
Length of output: 278
🏁 Script executed:
# Verify header coverage - check what other headers might be in axios.defaults.headers
rg "axios\.defaults\.headers|headers\s*=" src/client.ts -A 5 -B 2Repository: cloudglue/cloudglue-js
Length of output: 48
🏁 Script executed:
# Find documentation files
find . -maxdepth 2 -type f \( -name "README*" -o -name "CLAUDE*" -o -name "*.md" \)Repository: cloudglue/cloudglue-js
Length of output: 102
🏁 Script executed:
# Check client.ts for header setup
grep -n "headers\|Authorization\|x-sdk" src/client.ts | head -30Repository: cloudglue/cloudglue-js
Length of output: 369
🏁 Script executed:
# Look for any comments or documentation about why fetch is used for streaming
rg "fetch\(|stream\(" src/api/deep-search.api.ts -B 3 -A 1Repository: cloudglue/cloudglue-js
Length of output: 275
🏁 Script executed:
# Check if there are any TODOs or FIXMEs related to streaming or transport
rg -i "todo|fixme" src/api/deep-search.api.tsRepository: cloudglue/cloudglue-js
Length of output: 48
🏁 Script executed:
# Check CLAUDE.md for guidance on streaming or fetch usage
cat CLAUDE.mdRepository: cloudglue/cloudglue-js
Length of output: 3111
🏁 Script executed:
# Check README for any streaming or architecture guidance
cat README.md | head -100Repository: cloudglue/cloudglue-js
Length of output: 2732
Network-level errors during streaming are not wrapped in CloudglueError.
While HTTP response errors (non-ok status) are properly wrapped, connection failures, timeout errors, and stream-reading errors from fetch() will escape as raw errors instead of CloudglueError. For consistency with non-streaming endpoints, consider wrapping transport-level failures (e.g., around the fetch() call and stream consumption).
Note: SSE streaming requires a persistent connection incompatible with Zodios/Axios request-response design, so direct fetch() usage is appropriate here (similar to the documented exception for file uploads). The header extraction is adequate.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@src/api/deep-search.api.ts` around lines 186 - 223, The fetch call and
subsequent stream consumption in the deep-search endpoint can throw
transport-level errors that currently escape; wrap the entire network
interaction (the fetch(...) call and any downstream stream-reading logic) in a
try/catch and, on any thrown error, rethrow a CloudglueError so transport
failures are normalized with HTTP errors. Specifically, catch errors around the
fetch invocation and stream processing in the method containing the fetch, build
a CloudglueError using a sensible status (e.g., 0 or undefined), include
JSON.stringify(body) as the request payload, include headers (or an empty
object) from response.headers if available, and attach the original error
object/string as the error payload/responseData; ensure you still preserve
existing behavior for non-ok HTTP responses handled by the existing response.ok
branch.
kdr
added a commit
that referenced
this pull request
Jun 8, 2026
## Summary - Bumps the `cloudglue-api-spec` submodule to **v0.7.4** ([api-spec #98](cloudglue/cloudglue-api-spec#98)), which marks the include-gated Grain `source_metadata` fields as nullable. - Regenerates the client. The only generated change is in `GrainSourceMetadata` (`generated/common.ts`): `participants`, `highlights`, `ai_summary`, `ai_action_items`, `ai_template_sections`, and `calendar_event` move from `.optional()` → `.nullish()`, with matching `| null` TS types produced by the existing `generate.js` nullish/nullable transforms. - Bumps SDK version `0.7.12` → `0.7.13`. ## Verification - `npm run generate` is reproducible — re-running yields byte-identical output (only `common.ts` changes). - `npm run build` (tsc) passes. - Ran `getSourceMetadata()` against a live Grain connector: fetch + Zod parse both succeed with the new nullable schema. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Chores** * Version updated to 0.7.13 * Submodule reference updated <!-- end of auto-generated comment: release notes by coderabbit.ai -->
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
2 participants
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
client.deepSearchnamespace with sync, streaming (SSE), and background processing modes. IncludescreateDeepSearch,createStreamingDeepSearch,listDeepSearches,getDeepSearch,deleteDeepSearch,cancelDeepSearch, andwaitForReadypolling.client.dataConnectors.listFiles(connectorId, params)for browsing files in connected data sources (S3, GCS, Google Drive, Dropbox, Zoom) with provider-specific filtering and pagination.client.describe.updateDescribe(jobId, params)to toggleuse_in_default_index, anduse_in_default_indexoption oncreateDescribe.knowledge_basenow supports multiple source types (collections,files,default), addedtoolsfor function calling, andmax_output_tokensparameter.knowledge_baseis now optional.generate.jsto handle nullablePartial<{}>type mismatches andz.discriminatedUnionwith optional discriminators.Summary by CodeRabbit