Context Connectors #53

igor0 · 2025-12-15T20:53:51Z

An open-source library built on the Context Engine SDK that makes diverse sources searchable across agents and apps

Context Connectors enables users to:

Build indexes from Git repos (GitHub, GitLab, BitBucket), documentation websites, or local filesystem: index code, documentation, runbooks, schemas, configs, and more. Use DirectContext in the Context Engine SDK for custom sources.
Store indexes on a local filesystem for fast & simple access, or in S3 for persistent storage in production apps.
Search indexes via interactive agent, MCP for AI integrations, CLI for quick searches, or DirectContext in the Context Engine SDK for custom implementations.

Switch from openai() to openai.chat() to use the Chat Completions API instead of the Responses API. The Responses API is stateful and generates server-side IDs (fc_...) for function calls that are not persisted for Zero Data Retention (ZDR) organizations, causing multi-step tool calls to fail. The Chat Completions API is stateless and works correctly with ZDR.

- OpenAI: gpt-5.2 → gpt-5-mini - Anthropic: claude-sonnet-4-5 → claude-haiku-4-5 - Google: gemini-3-pro → gemini-3-flash-preview Also adds Phase 10 test results documenting: - ZDR compatibility fix (openai.chat vs openai) - Model availability testing - Multi-provider verification

- Add ./clients export path to package.json for programmatic API access - Export createMCPServer, runMCPServer, MCPServerConfig from clients module - Document Phase 11 programmatic API test results in test-results.md

Flip the default behavior: file tools (listFiles, readFile) are now enabled by default. Use --search-only to disable them. This is more intuitive - users get full functionality by default and explicitly opt out when they only want the search tool. - cmd-mcp: --search-only disables list_files/read_file tools - cmd-agent: --search-only disables listFiles/readFile tools - cmd-search: --search-only disables file access

…ansport - Add mcp-http-server.ts with runMCPHttpServer() and createMCPHttpServer() - Add mcp-serve CLI command with --port, --host, --cors, --base-path, --api-key options - Support API key authentication via Authorization: Bearer header - Support CORS for browser-based clients - Update README with HTTP server documentation and examples

Updated tool descriptions for search, list_files, and read_file to be more detailed and informative, adapting from Auggie CLI while keeping content appropriate for context-connectors: - Added multi-line descriptions with features and usage notes - Included condensed regex syntax guide for searchPattern - Clarified parameter semantics (1-based, inclusive, relative paths) - Removed coding-specific language to support general use cases Files modified: - src/clients/mcp-server.ts - src/clients/cli-agent.ts

…-specific store paths - Rename -k, --key flag to -n, --name across all CLI commands - Change default store location from CWD-relative .context-connectors to: - Linux: ~/.local/share/context-connectors (XDG Base Directory spec) - macOS: ~/Library/Application Support/context-connectors - Windows: %LOCALAPPDATA%\context-connectors - Add CONTEXT_CONNECTORS_STORE_PATH environment variable override - Priority order: --store-path CLI option > env var > platform default - Update README.md with new flag, Data Storage section, and env var docs

Replace importFromFile with temp file pattern with direct import() call. This is platform-neutral and avoids unnecessary filesystem operations.

Replace flat --source flag with subcommands for each source type: - index filesystem (alias: fs) - index github - index gitlab - index bitbucket - index website Each subcommand now shows only relevant options in --help. Extracted shared store options into reusable helper functions.

- Make --name optional for mcp and mcp-serve commands (default: all indexes) - Accept multiple index names with -n/--name <names...> - Discover available indexes at startup from store - Include available indexes in tool descriptions - Add index_name parameter to all tools (search, list_files, read_file) - Lazy initialization of SearchClient per index on first use - Cache initialized clients for reuse

Show source type, identifier, and relative sync time for each index: NAME SOURCE SYNCED augment-docs github://augmentcode/docs 1d ago lm-plot github://igor0/lm-plot 1d ago

- Change SourceMetadata to discriminated union with typed config per source - Store original ref (branch/tag) separately from resolvedRef (SHA) - Enable future re-indexing by preserving all source parameters - Add getSourceIdentifier() and getResolvedRef() helper functions - Maintain backward compatibility with legacy format indexes - Update all sources, consumers, and tests

- sync <name> updates a single index using stored config - sync --all updates all indexes - Keeps 'index' command name (clearer than 'add')

- Add explicit path format examples with ✅/❌ to prevent /repo confusion - Add example output for search results - Use consistent 'repository root' terminology - Add clearer parameter documentation with inline defaults - Improve regex guidance with specific unsupported pattern examples

The crawler used url.href (includes querystring) for de-duplication but url.pathname for storage paths. This caused URLs like /page?foo=1 and /page?bar=2 to both be crawled but stored to the same path, creating duplicates or overwrites. Now we use pathname consistently for both de-duplication and storage, ensuring each page is only crawled once regardless of query params.

The crawler uses url.href for de-duplication (different query params = different content, like browser caching). Now the storage path also includes a sanitized query string, so each unique URL gets its own file. Examples: - /articles?page=1 -> articles_page=1.md - /articles?page=2 -> articles_page=2.md - /search?q=cats -> search_q=cats.md This preserves paginated content and search results that vary by query.

igor0 · 2026-01-09T21:30:32Z

augment review

augment-app-staging

Review completed. 5 suggestions posted.

Comment augment review to trigger a new review at any time.

augment-app-staging · 2026-01-09T21:40:35Z

context-connectors/src/clients/mcp-http-server.ts

+    }
+
+    // Check if request is for MCP endpoint
+    if (!url.pathname.startsWith(basePath)) {


Using startsWith(basePath) can also match paths like /mcp2 when basePath is /mcp, potentially exposing the handler on unintended routes. Consider enforcing an exact match or a basePath/ boundary.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-09T21:40:35Z

context-connectors/src/clients/mcp-http-server.ts

+      }
+    } catch (error) {
+      if (!res.headersSent) {
+        res.writeHead(500, { "Content-Type": "application/json" });


Errors from parseBody (invalid JSON / body too large) currently fall into this catch and return a 500 JSON-RPC error. That makes client-side mistakes look like server failures and can cause noisy retries; consider mapping these to 4xx responses.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-09T21:40:35Z

context-connectors/src/bin/cmd-index.ts

+      const dir = options.saveContent;
+      await fs.mkdir(dir, { recursive: true });
+      for (const file of files) {
+        const filePath = path.join(dir, file.path);


When --save-content is used, path.join(dir, file.path) will honor .. segments or absolute paths inside file.path, which could write outside the target directory if the crawled site emits unexpected URLs. Consider normalizing/sanitizing file.path before using it as a filesystem path.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-09T21:40:35Z

context-connectors/src/bin/cmd-search.ts

+            indexKey = keyPath;
+          } else {
+            const prefix = keyPath.slice(0, lastSlash + 1);
+            indexKey = keyPath.slice(lastSlash + 1);


If the S3 URL ends with a trailing /, indexKey becomes an empty string here (keyPath.slice(lastSlash + 1)), which likely yields an invalid index name / object key. Consider validating keyPath doesn’t end with / (or trimming) before computing indexKey.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-09T21:40:35Z

context-connectors/src/core/indexer.ts

+    // Import previous context if available (for client-side deduplication),
+    // otherwise create a new one
+    const context = previousState
+      ? await DirectContext.import(previousState.contextState, {


Full reindex imports previousState.contextState into the context for deduplication, but fetchAll() may omit files that were deleted since the last run. If addToIndex is additive, those deleted files could persist in the index; consider whether full reindex should start from an empty context or explicitly remove missing paths.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

Using startsWith(basePath) could match unintended routes like /mcp2 when basePath is /mcp. Now we check for exact match OR subpath with a / boundary, preventing the handler from being exposed on unintended routes.

Errors from parseBody (invalid JSON, body too large) were falling through to the generic catch block and returning 500 Internal Server Error. Now we catch these explicitly and return appropriate status: - Invalid JSON: 400 Bad Request with JSON-RPC error -32700 - Body too large: 413 Payload Too Large with JSON-RPC error -32600 This prevents client-side mistakes from looking like server failures and avoids noisy retries.

When using --save-content, file.path from crawled URLs could contain .. segments or be absolute paths, potentially writing outside the target directory. Now we: 1. Normalize the path to resolve .. segments 2. Strip any leading .. sequences 3. Verify the resolved path stays within the target directory 4. Skip files with unsafe paths (with a warning)

If the S3 URL ends with a trailing slash, indexKey would become an empty string, yielding an invalid index name. Now we: 1. Trim trailing slashes from the key path 2. Validate that slashIdx is not -1 (handle s3://bucket without path) 3. Throw a clear error if the index key is missing

Instead of string matching on error messages to determine status codes, use a custom HttpError class with a statusCode property. This is more robust and idiomatic TypeScript.

igor0 · 2026-01-09T23:51:22Z

augment review

augment-app-staging

Review completed. 5 suggestions posted.

Comment augment review to trigger a new review at any time.

augment-app-staging · 2026-01-09T23:57:19Z

context-connectors/src/tools/list-files.ts

+  for (let i = 0; i < filteredEntries.length; i++) {
+    const entry = filteredEntries[i];
+    const entrySize = entry.path.length + entry.type.length + 5; // " [type]\n"
+    if (estimatedSize + entrySize > maxOutputLength) {


If maxOutputLength is smaller than the first entry’s estimated size, this returns an empty array, and formatListOutput will report “No files found” even though entries exist (they were just truncated). Callers also can’t tell truncation happened since truncated/omittedCount aren’t returned.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

I think that's OK

augment-app-staging · 2026-01-09T23:57:20Z

context-connectors/src/tools/read-file.ts

+    // Build output with gaps shown as "..."
+    const sortedLines = Array.from(lineNumbers).sort((a, b) => a - b);
+    const outputLines: string[] = [];
+    let lastLine = -2; // -2 so first line doesn't trigger gap


With lastLine = -2, the lineIdx > lastLine + 1 check is true for the first emitted line (e.g., 0 > -1), so search output will always start with ... despite the comment. This can imply leading content was omitted even when the first match/context line is at the top of the range.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-09T23:57:20Z

context-connectors/src/core/indexer.ts

+
+    // Add files to index
+    if (files.length > 0) {
+      await this.addToIndex(context, files);


When fullIndex reuses previousState for dedup, it only calls addToIndex(files) and never removes documents for files that no longer exist in the source, so deleted files may remain searchable while filesRemoved is always reported as 0. This seems especially likely on full-refresh paths like force-push handling.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

addressed... this was an over-optimization. when we are doing a full reindex (e.g., due to force push), we shouldn't try to reuse the existing index

augment-app-staging · 2026-01-09T23:57:20Z

context-connectors/src/stores/s3.ts

+  }
+
+  private getStateKey(key: string): string {
+    return `${this.prefix}${key}/${STATE_FILENAME}`;


S3 object keys are built via string concatenation (${this.prefix}${key}/…); if prefix is configured without a trailing / you’ll generate unexpected keys, and if key contains / it can break list()/delete() behavior with Delimiter: '/'. That can make some indexes undiscoverable or hard to delete depending on naming.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

addressed. prefix is now normalized to always end with /

augment-app-staging · 2026-01-09T23:57:20Z

context-connectors/src/clients/multi-index-runner.ts

+    const indexes: IndexInfo[] = [];
+    const validIndexNames: string[] = [];
+    for (const name of indexNames) {
+      const state = await store.loadSearch(name);


The comment says this loop “filter[s] out any that fail to load”, but any exception thrown by store.loadSearch(name) will abort initialization rather than skipping that index. A single corrupted/partially-written state could take down the whole multi-index runner.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

added try-catch to skip any indexes that caused a crash

…ine 1 With lastLine = -2, the check 'lineIdx > lastLine + 1' was true for the first emitted line at index 0 (0 > -1), so search output always started with '...' implying content was omitted even when the first match/context was at the top of the file. Change to -1 so only lines after index 0 trigger the gap marker.

When fetchChanges() returns null (e.g., force push), the previous code reused previousState for client-side deduplication. This caused deleted files to remain searchable since only addToIndex() was called, never removeFromIndex(). Fix: Always create a fresh context for full re-index. This ensures the index accurately reflects the current source state. Removed the unused previousState parameter from fullIndex() since it's now always null.

1. Normalize prefix to always end with '/' (unless empty) to prevent malformed keys like 'context-connectorsmykey/state.json' 2. Use sanitizeKey() to sanitize index keys, replacing '/' with '_' to prevent issues with S3 Delimiter '/' in list() that would make nested keys undiscoverable 3. Add guard in delete() to reject empty/unsafe keys that sanitize to empty string (consistent with FilesystemStore)

Wrap store.loadSearch() in try-catch so that corrupted or partially-written state files don't abort initialization of the entire multi-index runner. This matches the comment's stated intent of 'filtering out any that fail to load'.

igor0 · 2026-01-10T00:44:29Z

augment review

augment-app-staging

Review completed. 3 suggestions posted.

Comment augment review to trigger a new review at any time.

augment-app-staging · 2026-01-10T01:36:59Z

context-connectors/src/tools/list-files.ts

+    estimatedSize += entrySize;
+  }
+
+  // Add truncation info to last entry if needed (for display purposes)


listFiles() may truncate entries based on maxOutputLength, but it returns only FileInfo[] so callers can’t tell truncation occurred (and the comment implies comparing to a “reported total count”, which isn’t returned). Consider returning metadata (e.g., truncated/omittedCount/total) so consumers don’t assume the listing is complete.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-10T01:36:59Z

context-connectors/src/stores/s3.ts

+  }
+
+  async delete(key: string): Promise<void> {
+    // Guard against empty/unsafe keys


delete() rejects keys that sanitize to an empty string, but save()/loadState()/loadSearch() will still accept such keys and write objects that can’t be deleted via this API. Consider validating sanitizeKey(key) consistently across all operations to avoid creating undeletable state/search objects.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

addressed - added checks to the other functions

augment-app-staging · 2026-01-10T01:36:59Z

context-connectors/src/integrations/github-webhook-express.ts

+    next: NextFunction
+  ) {
+    try {
+      const signature = req.headers["x-hub-signature-256"] as string;


req.headers[...] can be string | string[] in Express/Node; casting to string here can yield an array at runtime and break signature/event validation. Consider normalizing/rejecting multi-valued headers explicitly (similar to the auth middleware in mcp-http-server.ts).

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

Changed listFiles() return type from Promise<FileInfo[]> to Promise<ListFilesResult> so callers can detect when output was truncated. The ListFilesResult includes: - entries: FileInfo[] - truncated?: boolean - omittedCount?: number Updated formatListOutput() to accept ListFilesResult and display a truncation notice when entries were omitted due to maxOutputLength.

Added validateAndSanitizeKey() helper that throws on empty sanitized keys. Used by getStateKey() and getSearchKey() which are called by all operations (save, loadState, loadSearch, delete). This prevents creating objects that cannot be deleted via this API.

Express/Node headers can be string | string[]. Previously the code cast them directly to string which could break signature/event validation at runtime. Now explicitly check for and reject array-valued headers, consistent with the auth middleware in mcp-http-server.ts.

igor0 · 2026-01-10T06:11:36Z

augment review

augment-app-staging

Review completed. 4 suggestions posted.

Comment augment review to trigger a new review at any time.

augment-app-staging · 2026-01-10T06:24:28Z

context-connectors/src/sources/github.ts

+      const { data } = await octokit.repos.getCommit({
+        owner: this.owner,
+        repo: this.repo,
+        ref: this.ref,


resolveRefToSha() calls repos.getCommit({ ref }), but ref can be the default "HEAD" (see constructor/CLI/docs) and GitHub’s commits API expects a branch/tag/SHA—"HEAD" is likely to 404 and make default indexing fail. Consider mapping "HEAD" to the repo’s default-branch head before calling getCommit.

Other Locations

context-connectors/src/sources/github.ts:111

context-connectors/src/bin/cmd-index.ts:67

context-connectors/README.md:98

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

Tested it and there is no problem: "my tests show that in practice GitHub's API does accept "HEAD" and resolves it correctly. This is likely because the Octokit library or GitHub's API internally resolves "HEAD" to the default branch."

augment-app-staging · 2026-01-10T06:24:28Z

context-connectors/src/sources/github.ts

+      const { data } = await octokit.repos.getContent({
+        owner: this.owner,
+        repo: this.repo,
+        path: directory,


When listing the repo root, this passes path: "" into repos.getContent(), which can become /contents/ and may 404 (the root contents endpoint is typically /contents without a {path}). This could make listFiles() fail for the root directory.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

This is not an issue:

Findings:
• The review comment is incorrect. The GitHub API and Octokit library handle path: "" correctly for listing root directory contents.
• The Octokit library properly constructs the URL, avoiding the trailing slash issue the reviewer was concerned about.
• The existing code works as intended.

augment-app-staging · 2026-01-10T06:24:28Z

context-connectors/src/stores/types.ts

+ * const store: IndexStore = new FilesystemStore();
+ *
+ * // Indexer uses full interface
+ * await store.save("my-project", indexState);


The docs example calls store.save("my-project", indexState), but IndexStore.save now requires (key, fullState, searchState); similarly, the FilesystemStore doc block still references store.load()/store.exists() APIs that no longer exist. This mismatch will likely confuse store implementers/users.

Other Locations

context-connectors/src/stores/filesystem.ts:24

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

updated the docs

augment-app-staging · 2026-01-10T06:24:28Z

context-connectors/src/tools/list-files.ts

+  }
+
+  const {
+    directory = "",


directory is passed through without normalization, so inputs like "/src/" or "./src" can produce double slashes or leading slashes and may break some Source.listFiles implementations. It might be worth normalizing directory similarly to readFile’s path handling.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

addressed - added normalization

Update code examples in types.ts and filesystem.ts to match the current API: - save() now requires (key, fullState, searchState) not (key, indexState) - load() no longer exists; use loadState() and loadSearch() instead - Added list() to the example for completeness

Add normalizePath() utility that removes leading './' and '/', trailing '/', and collapses multiple slashes. This prevents issues with Source.listFiles() implementations that may not handle paths like '/src/', './src', or 'src//lib' correctly.

igor0 · 2026-01-10T19:35:25Z

augment review

augment-app-staging

Review completed. 3 suggestions posted.

Comment augment review to trigger a new review at any time.

augment-app-staging · 2026-01-10T19:43:12Z

context-connectors/src/sources/github.ts

+   * Resolve ref (branch/tag/HEAD) to commit SHA
+   */
+  private async resolveRefToSha(): Promise<string> {
+    if (this.resolvedRef) {


resolveRefToSha() caches resolvedRef, which means a GitHubSource created with ref: "HEAD" (or a branch/tag) will keep using the first resolved SHA across later index() calls and may never observe new commits. Consider only caching when ref is already a commit SHA (or otherwise re-resolving per run).

Other Locations

context-connectors/src/sources/gitlab.ts:113

context-connectors/src/sources/bitbucket.ts:84

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

this is OK... the alternative would be the SHA changes across calls, which would confuse some callers. documented

augment-app-staging · 2026-01-10T19:43:12Z

context-connectors/src/integrations/github-webhook-express.ts

+      } else if (typeof req.body === "string") {
+        body = req.body;
+      } else {
+        body = JSON.stringify(req.body);


If req.body is an object (e.g. after express.json()), JSON.stringify(req.body) generally won’t match GitHub’s original raw payload bytes, so verifyWebhookSignature will consistently fail even with the correct secret. It may be safer to reject non-raw bodies and require express.raw() (or capture the raw body via the JSON middleware verify option) for signature validation.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

augment-app-staging · 2026-01-10T19:43:12Z

context-connectors/src/clients/mcp-http-server.ts

+          reject(new HttpError(`Request body too large (max ${MAX_BODY_SIZE} bytes)`, 413));
+          return;
+        }
+        body += chunk;


body += chunk implicitly decodes each Buffer chunk to UTF‑8 separately, which can corrupt multibyte characters split across chunks and lead to intermittent JSON parse failures. Collecting Buffers and decoding once (or calling req.setEncoding("utf8") and treating chunks as strings) avoids this.

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

Collect Buffer chunks and decode once at the end instead of using 'body += chunk' which implicitly decodes each chunk separately. This prevents corruption of multibyte UTF-8 characters that may be split across chunk boundaries.

The resolveRefToSha() method caches the resolved SHA for the lifetime of the instance to ensure consistency across multiple operations. Added documentation explaining this behavior and recommending users create a new source instance to pick up new commits.

JSON.stringify(req.body) won't match GitHub's original payload bytes (due to key ordering, whitespace, unicode escaping differences), so signature verification would consistently fail. Now requires raw body (Buffer from express.raw() or string) and returns 400 error with a helpful message if an object body is detected.

igor0 added 30 commits December 15, 2025 06:19

Context Connectors

96a5de5

.gitignore

50e9b3b

Remove AI SDK support

5d28d71

Test webhook

1490467

Fixes

5a6114e

Add ./clients export and MCP functions, document Phase 11 results

bf436a3

- Add ./clients export path to package.json for programmatic API access - Export createMCPServer, runMCPServer, MCPServerConfig from clients module - Document Phase 11 programmatic API test results in test-results.md

Remove test-results.md

7afb760

docs: add Phase 12 edge cases and error handling test results

8180d1b

Improvements

6185e23

Pagination fixes:

e057fb4

Fixes

f66fcc5

Remove unnecessary changes

3c1a44c

Tweak README.md

2c9abc0

Fix

4cc7f78

Use DirectContext.import() instead of temp file workaround

1109e41

Replace importFromFile with temp file pattern with direct import() call. This is platform-neutral and avoids unnecessary filesystem operations.

Improvements

aebdfd3

More fine-grained indexing updates to stdout

aed8199

feat: enhance list command with source metadata

df99b2c

Show source type, identifier, and relative sync time for each index: NAME SOURCE SYNCED augment-docs github://augmentcode/docs 1d ago lm-plot github://igor0/lm-plot 1d ago

feat: add sync command for updating indexes

1aeb63a

- sync <name> updates a single index using stored config - sync --all updates all indexes - Keeps 'index' command name (clearer than 'add')

Update comments

4546072

igor0 added 2 commits January 9, 2026 21:20

augment-app-staging bot reviewed Jan 9, 2026

View reviewed changes

igor0 added 5 commits January 9, 2026 21:45

Enforce exact match or subpath boundary for basePath

ac05f15

Using startsWith(basePath) could match unintended routes like /mcp2 when basePath is /mcp. Now we check for exact match OR subpath with a / boundary, preventing the handler from being exposed on unintended routes.

Use HttpError class for typed parseBody error handling

05b3dc8

Instead of string matching on error messages to determine status codes, use a custom HttpError class with a statusCode property. This is more robust and idiomatic TypeScript.

igor0 requested a review from richhankins January 9, 2026 23:51

augment-app-staging bot reviewed Jan 9, 2026

View reviewed changes

igor0 added 4 commits January 10, 2026 00:04

augment-app-staging bot reviewed Jan 10, 2026

View reviewed changes

igor0 added 3 commits January 10, 2026 05:43

augment-app-staging bot reviewed Jan 10, 2026

View reviewed changes

igor0 added 2 commits January 10, 2026 18:10

augment-app-staging bot reviewed Jan 10, 2026

View reviewed changes

igor0 added 3 commits January 10, 2026 20:56

Fix potential UTF-8 corruption when parsing request body

9284f36

Collect Buffer chunks and decode once at the end instead of using 'body += chunk' which implicitly decodes each chunk separately. This prevents corruption of multibyte UTF-8 characters that may be split across chunk boundaries.

Context Connectors #53

Are you sure you want to change the base?

Context Connectors #53

Uh oh!

Conversation

igor0 commented Dec 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

igor0 commented Jan 9, 2026

Uh oh!

augment-app-staging bot left a comment

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

igor0 commented Jan 9, 2026

Uh oh!

augment-app-staging bot left a comment

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

igor0 commented Jan 10, 2026

Uh oh!

augment-app-staging bot left a comment

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 10, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

augment-app-staging bot Jan 10, 2026

Choose a reason for hiding this comment

igor0 commented Dec 15, 2025 •

edited

Loading

igor0 Jan 10, 2026 •

edited

Loading