Seng/fix gh action #198
Seng/fix gh action #198
Conversation
WalkthroughThis update introduces a new "doc-qa" agent for question answering over developer documentation, adds supporting types and configuration, and modularizes the document sync process by replacing inline shell logic in GitHub Actions workflows with standalone scripts for file collection, validation, payload building, and webhook sending. Several TypeScript types and import paths are refactored for clarity and maintainability. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant doc-qa Agent
participant VectorStore
participant OpenAI GPT-4o
User->>doc-qa Agent: Sends question prompt
doc-qa Agent->>VectorStore: retrieveRelevantDocs(prompt)
VectorStore-->>doc-qa Agent: Returns relevant document vectors
doc-qa Agent->>VectorStore: retrieveDocumentBasedOnPath(path) (for each doc)
VectorStore-->>doc-qa Agent: Returns full document content(s)
doc-qa Agent->>OpenAI GPT-4o: Streams answer with system prompt and docs
OpenAI GPT-4o-->>doc-qa Agent: Streams answer JSON
doc-qa Agent-->>User: Returns streamed answer and doc paths
Possibly related PRs
Suggested reviewers
Poem
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
npm error Exit handler never called! 📜 Recent review detailsConfiguration used: CodeRabbit UI 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
⏰ Context from checks skipped due to timeout of 90000ms (1)
✨ Finishing Touches
🧪 Generate Unit Tests
🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
docs | b2c4f50 | Commit Preview URL | Jun 20 2025, 02:32 PM |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (3)
.github/workflows/sync-docs-full.yml (3)
3-5: Consider re-adding manual invocation
Might be useful to retainworkflow_dispatchfor emergency or ad-hoc runs:on: push: workflow_dispatch:
55-57: Include cleanup of temporary file list
mdx_files.txtremains after the run. Consider removing it as well to keep the workspace clean:+ rm -f mdx_files.txt
25-25: Remove trailing spaces
YAML lint reports trailing spaces on these blank lines. Please strip them to satisfy the linter.Also applies to: 40-40, 43-43, 54-54
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
.github/workflows/sync-docs-full.yml(2 hunks)
🧰 Additional context used
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs-full.yml
[error] 25-25: trailing spaces
(trailing-spaces)
[error] 40-40: trailing spaces
(trailing-spaces)
[error] 43-43: trailing spaces
(trailing-spaces)
[error] 54-54: trailing spaces
(trailing-spaces)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (4)
.github/workflows/sync-docs-full.yml (4)
3-5: Automatic trigger on push
Switching fromworkflow_dispatchtopushensures docs sync occurs on every commit without manual steps.
21-39: Iterative JSON assembly improves readability
Replacing a complexjqpipeline with a shell loop makes it easier to follow and handle file-level errors. Skips empty lines and non-existent paths gracefully.
41-42: Consistent removed files generation
Mirroring the changed-files approach to buildremoved_files.jsonkeeps the full-sync logic symmetric and clear.
44-53: Clear final payload construction
Using--slurpfilefor bothchangedandremovedarrays with a default fallback ensures the payload is always valid JSON.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
agent-docs/src/agents/doc-processing/docs-orchestrator.ts (1)
79-80: Verify the return type of vector upsert operation.The logging addition improves observability, but please verify that the
ctx.vector.upsert()method returns an object with alengthproperty. If the return type is not an array or array-like object, this will cause a runtime error.Run the following script to verify the upsert method signature and return type:
#!/bin/bash # Description: Search for the vector upsert method definition and its return type # Expected: Find the method signature and return type documentation # Search for upsert method definition in the codebase ast-grep --pattern 'upsert($_, $_) { $$$ }' # Search for upsert method type definitions rg -A 5 "upsert.*:" --type=ts # Search for vector context interface or type definitions rg -A 10 -B 2 "interface.*Vector|type.*Vector" --type=tsConsider the log level appropriateness:
Info-level logging for individual chunk operations might be verbose in production. Consider using debug level instead:
- ctx.logger.info('Upserted chunk: %o', result.length); + ctx.logger.debug('Upserted chunk: %o', result.length);
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
.github/workflows/sync-docs-full.yml(2 hunks)agent-docs/src/agents/doc-processing/docs-orchestrator.ts(1 hunks)
🧰 Additional context used
🧬 Code Graph Analysis (1)
agent-docs/src/agents/doc-processing/docs-orchestrator.ts (1)
agent-docs/src/agents/doc-processing/config.ts (1)
VECTOR_STORE_NAME(1-1)
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs-full.yml
[error] 25-25: trailing spaces
(trailing-spaces)
[error] 40-40: trailing spaces
(trailing-spaces)
[error] 43-43: trailing spaces
(trailing-spaces)
[error] 54-54: trailing spaces
(trailing-spaces)
[error] 73-73: no new line character at the end of file
(new-line-at-end-of-file)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
.github/workflows/sync-docs-full.yml (1)
25-25: Address the formatting issues that were previously flagged.The static analysis tools are still flagging the same trailing spaces and missing newline issues that were identified in previous reviews. These formatting issues need to be resolved.
Also applies to: 40-40, 43-43, 54-54, 57-57, 76-76
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (3)
.github/workflows/sync-docs-full.yml(2 hunks)agent-docs/agentuity.yaml(1 hunks)agent-docs/src/agents/doc-qa/index.ts(1 hunks)
✅ Files skipped from review due to trivial changes (1)
- agent-docs/agentuity.yaml
🧰 Additional context used
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs-full.yml
[error] 25-25: trailing spaces
(trailing-spaces)
[error] 40-40: trailing spaces
(trailing-spaces)
[error] 43-43: trailing spaces
(trailing-spaces)
[error] 54-54: trailing spaces
(trailing-spaces)
[error] 57-57: trailing spaces
(trailing-spaces)
[error] 76-76: no new line character at the end of file
(new-line-at-end-of-file)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (1)
.github/workflows/sync-docs-full.yml (1)
63-64: Good addition of authentication.The addition of the Authorization header with the bearer token from GitHub secrets properly secures the webhook endpoint communication.
Also applies to: 74-74
There was a problem hiding this comment.
Actionable comments posted: 4
♻️ Duplicate comments (2)
.github/workflows/sync-docs-full.yml (2)
16-16: Remove trailing spaces.
Eliminate trailing whitespace on this blank line to satisfy YAMLlint.
26-26: Add newline at end-of-file.
Include a trailing newline to comply with POSIX and YAML standards.
🧹 Nitpick comments (12)
bin/collect-all-files.sh (2)
15-23: Handle filenames with spaces and special characters
Using a line-based loop can break on filenames containing whitespace. Consider a null-delimited approach:-find content -type f -name "*.mdx" | \ - sed 's|^content/||' | \ - sort | \ - while read -r file; do +find content -type f -name "*.mdx" -print0 | \ + sed -z 's|content/||g' | \ + sort -z | \ + while IFS= read -r -d '' file; do
26-27: Avoid scanning twice
You’re callingfindagain to count files. Capture the list once and reuse it to improve performance:files=$(find content -type f -name "*.mdx") echo "$files" | sed ... file_count=$(echo "$files" | wc -l)bin/validate-files.sh (2)
27-31: Quote file paths in warnings
To avoid ambiguity when a filename contains spaces or special chars, wrap the path in quotes in warning messages:-echo "Warning: Unsafe path detected, skipping: $file" >&2 +echo "Warning: Unsafe path detected, skipping: '$file'" >&2
47-50: Graceful exit when no valid files
Exiting with an error if there are zero valid files can break incremental syncs with no changes. Consider exiting with status 0 and emitting an informational message instead:-if [ "$valid_count" -eq 0 ]; then - echo "Error: No valid files found" >&2 - exit 1 -fi +if [ "$valid_count" -eq 0 ]; then + echo "No valid files to process, exiting gracefully" >&2 + exit 0 +fibin/send-webhook.sh (2)
25-30: Stream payload directly tocurl
Storing the entire payload in a variable can be memory-intensive and may alter newlines. You can pipe stdin directly intocurl:- # Read payload from stdin - payload=$(cat) + # Stream payload from stdinThen build your
curl_argswith--data-binary @-and invoke:echo "$payload" | curl "${curl_args[@]}" "$WEBHOOK_URL"
33-36: Ensurejqis available
Before usingjqfor JSON validation, verify it’s installed to fail fast:command -v jq >/dev/null 2>&1 || { echo "Error: jq is required but not found" >&2; exit 1; }bin/build-payload.sh (2)
57-76: Use a JSON tool to build payload
Manually concatenating JSON risks quoting and escaping bugs. Consider usingjq:jq -Rn \ --arg repo "$REPO_NAME" \ --argjson changed "$(printf '%s\n' "${changed_files[@]}" | jq -R . | jq -s '.')" \ --argjson removed "$(printf '%s\n' "${removed_files[@]}" | jq -R . | jq -s '.')" \ '{repo: $repo, changed: $changed, removed: $removed}'
46-51: Verify mode consistency
Currently, infullmode you treat all changed files as removed but still include them in thechangedarray (so you delete and then re-upload). Confirm this is intentional; if you only want a purge, you may want to clearchanged_filesin full mode:if [ "$MODE" = "full" ]; then - removed_files=("${changed_files[@]}") + removed_files=("${changed_files[@]}") + changed_files=() echo "Full mode: treating all files as removed for refresh" >&2 fi.github/workflows/sync-docs-full.yml (1)
24-26: Simplify input handling and catch errors.
Replace the redundantcatwith direct input redirection and enablepipefailso any failure bubbles up immediately.run: | + set -o pipefail - cat all-files.txt | \ - ./bin/build-payload.sh "${{ github.repository }}" full | \ - ./bin/send-webhook.sh "https://agentuity.ai/webhook/f61d5ce9d6ed85695cc992c55ccdc2a6" "Bearer $AGENTUITY_TOKEN" + ./bin/build-payload.sh "${{ github.repository }}" full < all-files.txt | \ + ./bin/send-webhook.sh "https://agentuity.ai/webhook/f61d5ce9d6ed85695cc992c55ccdc2a6" "Bearer $AGENTUITY_TOKEN".github/workflows/sync-docs.yml (3)
29-31: Remove redundantcatand enforce error exit.
Streamline the payload step by readingchanged-files.txtdirectly and usingpipefailto catch failures without extra subshells.run: | + set -o pipefail - cat changed-files.txt | \ - ./bin/build-payload.sh "${{ github.repository }}" incremental | \ - ./bin/send-webhook.sh "https://agentuity.ai/webhook/f61d5ce9d6ed85695cc992c55ccdc2a6" "Bearer $AGENTUITY_TOKEN" + ./bin/build-payload.sh "${{ github.repository }}" incremental < changed-files.txt | \ + ./bin/send-webhook.sh "https://agentuity.ai/webhook/f61d5ce9d6ed85695cc992c55ccdc2a6" "Bearer $AGENTUITY_TOKEN"
21-21: Remove trailing spaces.
Clear any trailing whitespace here to resolve YAMLlint errors.
31-31: Add newline at end-of-file.
Ensure a final newline for POSIX and YAML compliance.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (7)
.github/workflows/sync-docs-full.yml(1 hunks).github/workflows/sync-docs.yml(1 hunks)bin/build-payload.sh(1 hunks)bin/collect-all-files.sh(1 hunks)bin/collect-changed-files.sh(1 hunks)bin/send-webhook.sh(1 hunks)bin/validate-files.sh(1 hunks)
🧰 Additional context used
🪛 GitHub Actions: Full Docs Sync to Vector Store
bin/collect-all-files.sh
[error] 20-20: Shell script error: 'echo' write error: Broken pipe. The script exited with code 1.
🪛 actionlint (1.7.7)
.github/workflows/sync-docs.yml
28-28: shellcheck reported issue in this script: SC2002:style:1:5: Useless cat. Consider 'cmd < file | ..' or 'cmd file | ..' instead
(shellcheck)
.github/workflows/sync-docs-full.yml
23-23: shellcheck reported issue in this script: SC2002:style:1:5: Useless cat. Consider 'cmd < file | ..' or 'cmd file | ..' instead
(shellcheck)
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs.yml
[error] 21-21: trailing spaces
(trailing-spaces)
[error] 31-31: no new line character at the end of file
(new-line-at-end-of-file)
.github/workflows/sync-docs-full.yml
[error] 16-16: trailing spaces
(trailing-spaces)
[error] 26-26: no new line character at the end of file
(new-line-at-end-of-file)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (4)
bin/send-webhook.sh (1)
1-4: Well-structured retry logic
The script cleanly handles retries with exponential backoff and error reporting.bin/collect-changed-files.sh (1)
1-55: Solid implementation for detecting changed MDX files
The script correctly parses added, modified, and deleted files with clear logging and safe Git refs validation.bin/build-payload.sh (1)
1-3: Clear usage and strict error handling
Shebang,set -euo pipefail, and the usage function are well-defined..github/workflows/sync-docs-full.yml (1)
4-4: Verify full-sync trigger scope.
The workflow now runs on every push across all branches, triggering a full vector sync each time. Confirm this broad trigger aligns with your performance and usage expectations.
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
.github/workflows/sync-docs-full.yml (1)
13-15: Enable fast-fail on collection pipeline
Includeset -o pipefailat the start of this step so any failure incollect-all-files.shorvalidate-files.shcauses the job to fail immediately.run: | + set -o pipefail ./bin/collect-all-files.sh | \ ./bin/validate-files.sh > all-files.txt
🧹 Nitpick comments (2)
.github/workflows/sync-docs-full.yml (2)
16-16: Remove trailing whitespace
This blank line contains trailing spaces and triggers a lint error.- +
26-26: Add newline at end of file
POSIX standards and linters expect a trailing newline.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
.github/workflows/sync-docs-full.yml(1 hunks)
🧰 Additional context used
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs-full.yml
[error] 16-16: trailing spaces
(trailing-spaces)
[error] 26-26: no new line character at the end of file
(new-line-at-end-of-file)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
.github/workflows/sync-docs-full.yml (1)
28-28: Secure webhook endpoint and add newline at EOF.
Replace the hard-coded URL with a GitHub secret and add a terminating newline:- ./bin/send-webhook.sh "https://agentuity.ai/webhook/…" "Bearer $AGENTUITY_TOKEN" + ./bin/send-webhook.sh "${{ secrets.SYNC_DOCS_WEBHOOK_URL }}" "Bearer $AGENTUITY_TOKEN" +# (ensure newline at end of file)
🧹 Nitpick comments (3)
.github/workflows/sync-docs-full.yml (1)
26-27: Avoid useless use ofcat.
ShellCheck SC2002 warns againstcat all-files.txt | cmd. Use input redirection instead:- cat all-files.txt | \ - ./bin/build-payload.sh "${{ github.repository }}" full | \ + ./bin/build-payload.sh "${{ github.repository }}" full < all-files.txt | \ ./bin/send-webhook.sh "https://agentuity.ai/webhook/…" "Bearer $AGENTUITY_TOKEN".github/workflows/sync-docs.yml (2)
33-34: Eliminate unnecessarycat.
Use input redirection to feedchanged-files.txt:- cat changed-files.txt | \ - ./bin/build-payload.sh "${{ github.repository }}" incremental | \ + ./bin/build-payload.sh "${{ github.repository }}" incremental < changed-files.txt | \ ./bin/send-webhook.sh "https://agentuity.ai/webhook/…" "Bearer $AGENTUITY_TOKEN"
24-24: Remove trailing whitespace.
YAMLlint flags trailing spaces on this line.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (3)
.github/workflows/sync-docs-full.yml(1 hunks).github/workflows/sync-docs.yml(1 hunks)bin/collect-all-files.sh(1 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
- bin/collect-all-files.sh
🧰 Additional context used
🪛 actionlint (1.7.7)
.github/workflows/sync-docs-full.yml
24-24: shellcheck reported issue in this script: SC2002:style:2:5: Useless cat. Consider 'cmd < file | ..' or 'cmd file | ..' instead
(shellcheck)
.github/workflows/sync-docs.yml
31-31: shellcheck reported issue in this script: SC2002:style:2:5: Useless cat. Consider 'cmd < file | ..' or 'cmd file | ..' instead
(shellcheck)
🪛 YAMLlint (1.37.1)
.github/workflows/sync-docs-full.yml
[error] 17-17: trailing spaces
(trailing-spaces)
[error] 28-28: no new line character at the end of file
(new-line-at-end-of-file)
.github/workflows/sync-docs.yml
[error] 24-24: trailing spaces
(trailing-spaces)
[error] 35-35: no new line character at the end of file
(new-line-at-end-of-file)
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (7)
.github/workflows/sync-docs-full.yml (3)
12-12: Step definition looks good.
This clearly separates file collection and validation into its own job step.
14-16: Error handling withset -euo pipefailis correctly implemented.
It ensures failures in either script abort the workflow immediately.
18-19: Clarity of output is adequate.
Logging the list of files to sync aids debugging..github/workflows/sync-docs.yml (4)
15-16: Full-history checkout added.
fetch-depth: 0ensures diffs include the full commit history needed for comparisons.
20-23: Pipeline error handling & fetch context are solid.
Combiningset -euo pipefailwithgit fetch originprovides robust failure detection and proper diff base.
26-27: Logging of changed files is clear.
Echoing and displayingchanged-files.txtaids troubleshooting.
31-31: Repeated error handling is consistent.
Reapplyingset -euo pipefailin the second step is good practice.
| set -euo pipefail | ||
| cat changed-files.txt | \ | ||
| ./bin/build-payload.sh "${{ github.repository }}" incremental | \ | ||
| ./bin/send-webhook.sh "https://agentuity.ai/webhook/f61d5ce9d6ed85695cc992c55ccdc2a6" "Bearer $AGENTUITY_TOKEN" |
There was a problem hiding this comment.
Secure webhook URL and add newline at EOF.
Replace the hard-coded endpoint with a GitHub secret and append a newline:
- ./bin/send-webhook.sh "https://agentuity.ai/webhook/…"
+ ./bin/send-webhook.sh "${{ secrets.SYNC_DOCS_WEBHOOK_URL }}"
+# (ensure newline at end of file)Committable suggestion skipped: line range outside the PR's diff.
🧰 Tools
🪛 YAMLlint (1.37.1)
[error] 35-35: no new line character at the end of file
(new-line-at-end-of-file)
🤖 Prompt for AI Agents
In .github/workflows/sync-docs.yml at line 35, the webhook URL is hard-coded,
which is insecure. Replace the URL string with a reference to a GitHub secret
(e.g., ${{ secrets.WEBHOOK_URL }}) to secure the endpoint. Also, ensure the file
ends with a newline character to comply with POSIX standards.
There was a problem hiding this comment.
Actionable comments posted: 5
🧹 Nitpick comments (1)
agent-docs/src/agents/doc-qa/index.ts (1)
89-95: Clarify the empty space query and arbitrary limit.The query uses a single space
' 'and a limit of 10000, which seems arbitrary and potentially inefficient for large document collections.Consider using a more explicit approach:
async function retrieveDocumentBasedOnPath(ctx: AgentContext, path: string): Promise<string> { const dbQuery = { - query: ' ', - limit: 10000, + query: '', // Empty query to match all chunks for this path + limit: 1000, // More reasonable limit, adjust based on your max document size metadata: { path: path } }Or better yet, if the vector store supports it, use a metadata-only query without requiring a text query.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (7)
agent-docs/src/agents/doc-processing/docs-orchestrator.ts(2 hunks)agent-docs/src/agents/doc-processing/docs-processor.ts(1 hunks)agent-docs/src/agents/doc-processing/embed-chunks.ts(1 hunks)agent-docs/src/agents/doc-processing/types.ts(1 hunks)agent-docs/src/agents/doc-qa/config.ts(1 hunks)agent-docs/src/agents/doc-qa/index.ts(1 hunks)agent-docs/src/agents/doc-qa/types.ts(1 hunks)
✅ Files skipped from review due to trivial changes (5)
- agent-docs/src/agents/doc-processing/embed-chunks.ts
- agent-docs/src/agents/doc-processing/types.ts
- agent-docs/src/agents/doc-processing/docs-processor.ts
- agent-docs/src/agents/doc-qa/types.ts
- agent-docs/src/agents/doc-qa/config.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- agent-docs/src/agents/doc-processing/docs-orchestrator.ts
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (2)
agent-docs/src/agents/doc-qa/index.ts (2)
1-7: LGTM: Clean imports and proper typing.The imports are well-organized and include proper TypeScript types.
88-116: Good error handling implementation.The error handling in
retrieveDocumentBasedOnPathis well-implemented with proper logging and graceful degradation.
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
agent-docs/src/agents/doc-qa/index.ts (1)
25-39: JSON response format is requested but not validated or parsed.The system prompt explicitly requests a JSON response with "answer" and "documents" fields, but the streaming response bypasses any JSON validation. This could lead to malformed responses being sent to clients if the LLM doesn't follow the JSON format correctly.
🧹 Nitpick comments (3)
agent-docs/src/agents/doc-qa/index.ts (3)
58-60: Clean up formatting and variable declaration.The empty lines and
vardeclaration should be cleaned up for consistency.try { - - - var vectors = await ctx.vector.search(VECTOR_STORE_NAME, dbQuery); + const vectors = await ctx.vector.search(VECTOR_STORE_NAME, dbQuery);
93-94: Consider using a more explicit query for document retrieval.Using a single space
' 'as the query parameter seems unusual. Consider using an empty string''or a more explicit approach like omitting the query parameter entirely when filtering by metadata only.const dbQuery = { - query: ' ', + query: '', limit: 10000, metadata: { path: path } }
104-108: Optimize redundant type casting.The metadata is cast to
ChunkMetadatatwice in consecutive lines, which is redundant and can be simplified.const sortedVectors = vectors .map(vec => { + const metadata = vec.metadata as ChunkMetadata; + return { + metadata, + index: metadata.chunkIndex + }; + }) - .map(vec => ({ - metadata: vec.metadata as ChunkMetadata, - index: (vec.metadata as ChunkMetadata).chunkIndex - })) .sort((a, b) => a.index - b.index);
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
agent-docs/src/agents/doc-qa/index.ts(1 hunks)
🧰 Additional context used
🧠 Learnings (1)
agent-docs/src/agents/doc-qa/index.ts (1)
Learnt from: afterrburn
PR: agentuity/docs#198
File: agent-docs/src/agents/doc-qa/index.ts:9-50
Timestamp: 2025-06-20T13:09:14.197Z
Learning: In the agent-docs codebase, the user prefers to let exceptions throw from Agent functions rather than adding local try-catch blocks, allowing errors to be handled at a higher level in the application.
⏰ Context from checks skipped due to timeout of 90000ms (1)
- GitHub Check: Workers Builds: docs
🔇 Additional comments (1)
agent-docs/src/agents/doc-qa/index.ts (1)
94-94: Verify the high limit for document chunk retrieval.The limit of 10000 for retrieving document chunks seems very high. This could potentially cause performance issues or memory constraints for large documents.
Consider whether this limit is necessary or if it could be reduced to a more reasonable value based on typical document sizes in your system.
…into seng/fix-gh-action
| model: openai('gpt-4o'), | ||
| system: systemPrompt, | ||
| prompt: prompt, | ||
| maxTokens: 2048, |
There was a problem hiding this comment.
This model's size is 128,000 tokens - why so low here?
There was a problem hiding this comment.
Oh yeah I'll extend the size in the next PR. Mainly, I wanted to keep the generation concise and short for now. Will push its limit.
Summary by CodeRabbit