viberag 0.4.1 → 0.4.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +26 -21
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -51,27 +51,10 @@ When using a coding agent like [Claude Code](https://claude.ai/code), add `use v
51
51
  - **Semantic code search** - Find code by meaning, not just keywords
52
52
  - **Flexible embeddings** - Local model (offline, free) or cloud providers (Gemini, Mistral, OpenAI)
53
53
  - **MCP server** - Works with Claude Code, Cursor, VS Code Copilot, and more
54
- - **Automatic Incremental indexing** - Watches for file changes (respects `.gitignore`) and reindexes only what has changed in real time
54
+ - **Automatic incremental indexing** - Watches for file changes (respects `.gitignore`) and reindexes only what has changed in real time
55
55
  - **Resilient indexing** - Retries embedding errors and reports failed batches in `/status`
56
56
  - **Multi-language support** - TypeScript, JavaScript, Python, Go, Rust, and more
57
-
58
- ## Troubleshooting
59
-
60
- ### Watcher EMFILE (too many open files)
61
-
62
- Large repos can exceed OS watch limits. The watcher now honors `.gitignore`, but if you still see EMFILE:
63
-
64
- - Add more ignores in `.gitignore` to reduce watched files.
65
- - Increase OS limits:
66
- - macOS: raise `kern.maxfiles`, `kern.maxfilesperproc`, and `ulimit -n`
67
- - Linux: raise `fs.inotify.max_user_watches`, `fs.inotify.max_user_instances`, and `ulimit -n`
68
-
69
- ### Index failures (network/API errors)
70
-
71
- Embedding batches retry up to 10 attempts. If failures persist:
72
-
73
- - Run `/status` to see failed batch counts.
74
- - Re-run `/index` to retry failed files once connectivity is stable.
57
+ - **Blazing fast** - The data storage and search functionality is local on your machine, meaning the full power of your machine can churn through massive amounts of data and execute complex search queries in milliseconds.
75
58
 
76
59
  ### How It Works:
77
60
 
@@ -87,9 +70,13 @@ Semantic search is especially useful in monorepos, where you may be trying to un
87
70
 
88
71
  ### Embedding Models
89
72
 
90
- - You can use a locally run embedding model ([Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B)) so that nothing leaves your machine.
73
+ _All options store embeddings and indexed data on your local machine_
74
+
75
+ - **Local:** You can use a locally run embedding model ([Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B)) so that nothing leaves your machine. This has a smaller vocabulary and is only recommended for privacy and offline concerns.
91
76
 
92
- - SOTA API based embeddings from [Gemini](https://ai.google.dev/gemini-api/docs/embeddings), [OpenAI](https://platform.openai.com/docs/guides/embeddings), and [Mistral](https://docs.mistral.ai/capabilities/embeddings) are also supported.
77
+ - **Recommended:** API generated embeddings from [Gemini](https://ai.google.dev/gemini-api/docs/embeddings), [OpenAI](https://platform.openai.com/docs/guides/embeddings), and [Mistral](https://docs.mistral.ai/capabilities/embeddings) are recommended for the largest vocabulary and highest quality semantic meaning.
78
+ - These embeddings are very affordable at ~10 - 15 cents per million tokens.
79
+ - A typical codebase can be indexed for pennies
93
80
 
94
81
  ## MCP Server
95
82
 
@@ -724,3 +711,21 @@ Use `codebase_parallel_search` to run multiple search strategies in a single cal
724
711
  ```
725
712
 
726
713
  This provides comprehensive coverage without multiple round-trips.
714
+
715
+ ## Troubleshooting
716
+
717
+ ### Watcher EMFILE (too many open files)
718
+
719
+ Large repos can exceed OS watch limits. The watcher now honors `.gitignore`, but if you still see EMFILE:
720
+
721
+ - Add more ignores in `.gitignore` to reduce watched files.
722
+ - Increase OS limits:
723
+ - macOS: raise `kern.maxfiles`, `kern.maxfilesperproc`, and `ulimit -n`
724
+ - Linux: raise `fs.inotify.max_user_watches`, `fs.inotify.max_user_instances`, and `ulimit -n`
725
+
726
+ ### Index failures (network/API errors)
727
+
728
+ Embedding batches retry up to 10 attempts. If failures persist:
729
+
730
+ - Run `/status` to see failed batch counts.
731
+ - Re-run `/index` to retry failed files once connectivity is stable.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "viberag",
3
- "version": "0.4.1",
3
+ "version": "0.4.2",
4
4
  "description": "Local code RAG for AI coding assistants - semantic search via MCP server",
5
5
  "license": "AGPL-3.0",
6
6
  "keywords": [