AutoDev Catalyser #30
Labels
enhancement
New feature or request
Comments
phodal
added a commit
that referenced
this issue
May 14, 2024
Add Catalyser class for semantic code search functionality.
phodal
added a commit
that referenced
this issue
May 14, 2024
Refactor the retrieveContextItems function in DefaultRetrieval.ts for clarity and consistency.
phodal
added a commit
that referenced
this issue
May 14, 2024
- Update method name from getHydeTemplate to renderHydeTemplate for clarity.
phodal
added a commit
that referenced
this issue
May 14, 2024
- Add evaluation step to prompt processing for keyword analysis.
phodal
added a commit
that referenced
this issue
May 14, 2024
Added console.log statements for search results and evaluation step in Catalyser.ts for better debugging and visibility.
phodal
added a commit
that referenced
this issue
May 14, 2024
Added functionality to request user input in the sidebar webview. This commit enhances the extension by allowing the sidebar webview to request user input, improving user interaction.
phodal
added a commit
that referenced
this issue
May 14, 2024
- Update class and method names from RankedKeywords to QuestionKeywords - Rename file from RankedKeywords.ts to QuestionKeywords.ts - Adjust references and tests accordingly.
phodal
added a commit
that referenced
this issue
May 15, 2024
- Update method names in LlmProvider for clarity and consistency. - Replace `instance()` with `codeCompletion()` and `chatCompletion()`.
phodal
added a commit
that referenced
this issue
May 15, 2024
Add enum for LLM strategy options and update code and chat completion methods to accept a strategy parameter. Default strategy set for each method.
phodal
added a commit
that referenced
this issue
May 15, 2024
Moved keyword context interfaces from the prompt management file to the search strategy file for better organization and cohesion.
phodal
added a commit
that referenced
this issue
May 15, 2024
- Refactor HydeKeywordsStrategy and Catalyser for optimization.
phodal
added a commit
that referenced
this issue
May 15, 2024
Added support for text ranges in code search to improve code retrieval accuracy and display.
phodal
added a commit
that referenced
this issue
May 15, 2024
- Update search strategy steps to include 'Retrieve' step in HydeKeywordsStrategy.ts and HydeStep.ts.
phodal
added a commit
that referenced
this issue
May 15, 2024
Refactor step handling in HydeKeywordsStrategy.ts for better clarity and maintainability.
phodal
added a commit
that referenced
this issue
May 15, 2024
- Added logging for the instruction parameter in the HydeKeywordsStrategy file to aid debugging and monitoring.
phodal
added a commit
that referenced
this issue
May 15, 2024
…#30 Refactor query term creation in HydeKeywordsStrategy to improve readability and maintainability.
phodal
added a commit
that referenced
this issue
May 15, 2024
The `HydeKeywordsStrategy` class is refactored to prioritize text code in a specific order: current document, recently documents, and all documents. This improves the generation of keywords from a query and the retrieval of similar code by symbols.
phodal
added a commit
that referenced
this issue
May 15, 2024
- Refactored code chunking and indexing logic for better performance and readability. - Updated the `Chunker` classes to include the `language` property in the generated chunks. - Added the `languageFromPath` function to determine the language based on the file extension. - Optimized the `BasicChunker` and `ConstructCodeChunker` to pass the `language` parameter to the generated chunks. - Updated the database schema and queries to include the `language` column. - Updated the test cases to reflect the changes in the code chunking process.
phodal
added a commit
that referenced
this issue
May 15, 2024
phodal
added a commit
that referenced
this issue
May 15, 2024
phodal
added a commit
that referenced
this issue
May 15, 2024
Refactor the code for handling embeddings in the LanceDbIndex file. Instead of directly assigning the result of the embeddingsProvider.embed() method to the embeddings variable, wrap it in a try-catch block. This change allows for better error handling and logging when embedding chunks fails.
phodal
added a commit
that referenced
this issue
May 15, 2024
phodal
added a commit
that referenced
this issue
May 15, 2024
This commit adds support for Tree-sitter language tags in the code search feature. It includes the addition of language-specific tag schemas for various languages such as QL, Go, Elm, C++, C, Elixir, Java, PHP, Python, Rust, Ruby, OCaml, and TypeScript. These schemas define the structure and naming conventions for different code elements like classes, functions, methods, modules, and more. Additionally, the commit also modifies the CodeSnippetsCodebaseIndex and LanceDbIndex files to import the necessary dependencies and update file paths.
phodal
added a commit
that referenced
this issue
May 15, 2024
phodal
added a commit
that referenced
this issue
May 16, 2024
…30 Add support for generating code snippets that can be returned as answers to code search engine queries. The snippets should be written in a programming or markup language likely given the query and should be between 5 and 10 lines long. The snippets are surrounded by triple backticks for proper formatting.
phodal
added a commit
that referenced
this issue
May 16, 2024
Add support for generating synthetic documents based on user input to improve semantic search recall. This includes the implementation of the `HydeCodeStrategy` class, which generates synthetic documents based on the query and extracts code. The strategy first tries semantic search and falls back to code search if there are no or few results. Additionally, the code includes the extraction of the Chunk with `NamedElement` and `extra_chunks` to a class or function. Related files: - `src/code-search/search-strategy/HydeCodeStrategy.ts` - `src/test/codesearch/strategy/RankedKeywords.test.ts`
phodal
added a commit
that referenced
this issue
May 16, 2024
Add logging statements to print the keywords and chunks during the code search process.
phodal
added a commit
that referenced
this issue
May 16, 2024
- Added a new optional `language` parameter to the `retrieveChunks` and `retrieveContextItems` functions in `HydeKeywordsStrategy.ts` and `DefaultRetrieval.ts` respectively. - The `language` parameter allows for filtering results based on the specified programming language. - Modified the SQL query in `FullTextSearch.ts` to include the `language` parameter in the WHERE clause when retrieving chunks.
phodal
added a commit
that referenced
this issue
May 20, 2024
- Add support for creating a new session in the autoDev commands.
phodal
added a commit
that referenced
this issue
May 20, 2024
Ensure proper handling of empty code snippets in search strategies. - Add checks for empty code chunks and provide appropriate feedback. - Display a message when failing to embed a query or chunk.
phodal
added a commit
that referenced
this issue
May 20, 2024
Ensure proper handling when no ranges are available to avoid errors.
phodal
added a commit
that referenced
this issue
May 20, 2024
- Update the UUID generation method to use the `crypto.randomUUID()` function for better security and uniqueness.
phodal
added a commit
that referenced
this issue
May 20, 2024
Add a session check before initializing the LocalEmbeddingProvider to prevent redundant initialization.
phodal
added a commit
that referenced
this issue
May 21, 2024
Added methods to retrieve git history and changes by commit hash in the Retrieval and GitAction classes. This allows for indexing and searching relative to commit history.
phodal
added a commit
that referenced
this issue
May 21, 2024
Added an option to retrieve Git changes in the code search functionality. This includes changes to the `DefaultRetrieval.ts`, `IdeAction.ts`, and `Retrieval.ts` files. The new feature allows for the retrieval of Git changes based on commit messages using the TfIdfSemanticChunkSearch.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Added `withGitChange` option to `HydeKeywordsStrategy` and `HydeCodeStrategy`. - Changed return type of `search` method in `TfIdfSemanticChunkSearch` from `string[]` to `number[]`. - Updated `retrieveGit` method in `Retrieval` to include a threshold parameter and improved commit retrieval logic. - Reduced the default number of commits retrieved in `getHistoryCommits` method in `GitAction`. - Updated `diffWith` method in `GitAction` to return a joined string of changes. - Replaced `stopwords` with `ourStopwords` in `Tfidf` to allow for custom stopwords.
phodal
added a commit
that referenced
this issue
May 21, 2024
…tions #30 Increased the maximum number of history commits from 50 to 500 in `getHistoryCommits` method. Added type annotations to `diffResult` and `diffs` variables in `getRepositoryChanges` and `getChangeByHash` methods respectively. Also, added a comment to fix `getChangeByHashInRepo` method in the future as the official API doesn't support this feature yet.
phodal
added a commit
that referenced
this issue
May 21, 2024
…ements #30 The GitAction class in the editor-api module has been significantly enhanced. A new method `exec` has been added to execute git commands. The method `_getRepo` has been refactored to use the new `gitApi` promise. The `constructor` has been updated to initialize `gitApi` and `gitExecutablePath`. Additionally, copyright notice has been added at the beginning of the file.
phodal
added a commit
that referenced
this issue
May 21, 2024
…fIdfSemanticChunkSearch #30 Replaced 'withGitChange' option with 'withCommitMessageSearch' in HydeKeywordsStrategy, HydeCodeStrategy, and DefaultRetrieval. Renamed 'TfIdfSemanticChunkSearch' to 'TfIdfChunkSearch' across multiple files. Added error handling for git retrieval in DefaultRetrieval. Updated test cases to reflect these changes.
phodal
added a commit
that referenced
this issue
May 21, 2024
The commit message search feature has been enabled in HydeKeywordsStrategy. The retrieval process in DefaultRetrieval has been updated to accommodate this change. The number of retrievals is reduced when searching commit messages due to their length. A TODO has been added in GitAction to split show by multiple content with git changes.
phodal
added a commit
that referenced
this issue
May 21, 2024
Added a new GitParser module to the editor-api that can parse Git logs into a structured format. This includes the status of the file (added, modified, deleted) and the changes made. Also updated the GitAction module to use this new parser when retrieving changes by hash. Added corresponding tests to verify the correct functionality of the parser.
phodal
added a commit
that referenced
this issue
May 21, 2024
Add LlmReranker prompt support to PromptManager and templates.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Update LLMReranker constructor to use OpenAICompletion. - Refactor complete method in OpenAICompletion for better abstraction.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Update embedding provider instantiation to use non-null assertion operator. - Export `OllamaEmbeddingsProvider` class instead of default export. - Add a todo comment to load configuration for creating an embedding provider.
phodal
added a commit
that referenced
this issue
May 21, 2024
Refactor the embedding provider management by introducing a new `EmbeddingsProviderManager` namespace. This change includes renaming `LocalEmbeddingProvider` to `LocalEmbeddingsProvider`, updating references, and centralizing provider creation and initialization logic.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Add support for multiple provider types including Local, OpenAI, and Ollama.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Update ParsedFileChange interface to include content field. - Refactor GitParser and GitAction to handle content changes. - Return ParsedFileChange array instead of concatenated string.
phodal
added a commit
that referenced
this issue
May 21, 2024
A new method, `pathSimilarity`, has been added to the `JaccardSimilarity` class. This method calculates the similarity score between a path and a set of strings. Corresponding test case has also been added to ensure the correct functionality of the method.
phodal
added a commit
that referenced
this issue
May 21, 2024
The commit introduces the Jaccard similarity algorithm to improve the search results in the code search feature. It filters out results with a similarity score less than 0.5 to ensure more relevant search results.
phodal
added a commit
that referenced
this issue
May 21, 2024
Refactor Jaccard similarity calculation method for efficiency and clarity.
phodal
added a commit
that referenced
this issue
May 21, 2024
Renamed DomainTerm to TeamTerm and updated related services and files.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Added support for custom team terms with a new CSV file and updated service to handle custom prompts directory. - Updated TeamTerm interface to include clearer documentation for term and language fields. - Refactored SettingService to handle legacy config migration for custom prompts directory.
phodal
added a commit
that referenced
this issue
May 21, 2024
Added new terms to the team terms CSV file and improved the TeamTermService by adding a singleton pattern and fetch method for retrieving terms.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Added QueryExpansion class to handle query expansion by replacing terms with their localized versions along with term identifiers.
phodal
added a commit
that referenced
this issue
May 21, 2024
- Refactored query expansion logic to enhance performance and accuracy.
phodal
added a commit
that referenced
this issue
May 22, 2024
Refactored TreeSitterFileManager to improve cache handling and update mechanism. Removed unnecessary imports and adjusted the usage of TreeSitterFileManager in AutoDevCodeLensProvider. Also, updated the pub-sub mechanism in TreeSitterFile to support one-time listeners.
phodal
added a commit
that referenced
this issue
May 22, 2024
The Map used for caching in TreeSitterFileManager has been replaced with an LRUCache. This change should improve the performance and efficiency of the cache management.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Validations
Problem
A catalysis for coding
Solution
No response
Todos
The text was updated successfully, but these errors were encountered: