$_ Concave
Github

Open Source, Trusted

Autonomous Software Engineering

Goal for concave
We are creating Open-Source, Trusted, and Fully Autonomous Software Engineering.
We design and develop Concave for existing LLMs to lower the barrier to participating in autonomous software engineering.
Introducing Concave first product
Concave L·IDE
World First IDE for LLMs
- Provider the most powerful and efficient tools for LLMs
- Hybrid Code Search. help LLMs to understand the codebase.
- Full-Text Search. (Local)
- Source Code
- Git Info. Commit, Issues, and PRs ... [WIP]
- Symbols Search. ( TiDB Serverless)
- Faster Pre-Index, [Code Intelligence]
- Symbols: methods, class, variables...
- Definitions, References, and more...
- Semantic Search. ( TiDB Vector Search)
- Symbols, Code snippet
- Documents
- Provider the code workspace for LLMs
L·IDE Hybrid Code Search
Design for helping LLMs to understand the codebase.
concave-workspace@pytest#4787fd
$concave key_search 'EncodedFile'
/src/_pytest/capture.py 401 | class EncodedFile(io.TextIOWrapper): 530 | tmpfile = EncodedFile( /testing/test_capture.py 1479 | ef = capture.EncodedFile(tmpfile, encoding="utf-8")
$concave semantic_search 'EncodedFile'
connecting to tidb serverless snippet/pytest-dev/pytest/4787fd... q: Code related to 'EncodedFile' -----------top 6-------------- 1 | 0.744 _pytest.capture.EncodedFile 2 | 0.737 _pytest.capture.DontReadFromInput.encoding 3 | 0.737 _pytest.capture.EncodedFile.__slots__ 4 | 0.733 _pytest.capture.EncodedFile.name 5 | 0.723 _pytest.pytester.Testdir._makefile 6 | 0.716 _pytest.capture.EncodedFile.mode ------------------------------ model: voyage-code-2, _dimension: 1536
$concave symbol_search 'EncodedFile'
connecting to tidb serverless symbols/pytest-dev/pytest/4787fd... _pytest.capture.EncodedFile# kind: CLASS Defined: /src/_pytest/capture.py#425 --------------------------------------------------------- 425 | class EncodedFile(object): 426 | errors = "strict" # possibly needed by py3 code (issue555) 427 | 428 | def __init__(self, buffer, encoding): 429 | self.buffer = buffer ...... References: - src/_pytest/capture.py#422 @_pytest.capture.safe_text_dupfile() --------------------------------------------------------- 420 | mode += "b" 421 | f = os.fdopen(newfd, mode, 0) # no buffering 422 | return EncodedFile(f, encoding or default_encoding)
$concave history_search 'EncodedFile'
Latest 5 changes: - 698c4e7, capture: track current state in _state attributes, 2019-05-14 14:54:03 - 08ded29, capture: do not set logging.raiseExceptions = False, 2019-04-10 14:07:29 - 951213e, Use new suspend/resume in global_and_fixture_disabled, 2019-03-17 17:27:25 - d406786, pdb: handle capturing with fixtures only, 2019-03-18 17:27:59 - da81c1e, Merge pull request #4950 from blueyed/capture, 2019-03-19 17:12:22
Concave Fleet
Experimental, end-to-end issue auto-resolution solution.
Based on Concave LIDE + TiDB Serverless.
Significantly outperforms existing open-source solutions.

╔═══════════════════════════════════════════════════════════════════════════════════╗
║                                                                                   ║        
║  Concave Fleet GPT4o | 78% | 89% | ███████████████████████████████████████▒▒▒▒▒   ║
║                                                                                   ║
║   SWEAgent Calude3   | 21% | 42% | ██████████▒▒▒▒▒▒▒▒▒▒▒                          ║
║   SWEAgent Calude3.5 | 42% | 47% | █████████████████████▒▒▒                       ║
║   SWEAgent GPT4o     | 36% | 64% | ██████████████████▒▒▒▒▒▒▒▒▒▒▒▒▒▒               ║
║                                                                                   ║
║   Identify Accuracy  | File | Symbol |                                            ║
║                                                                                   ║
║                                                                                   ║
║  SWE_Bench Verified, filter:repo=PyTest                                           ║ 
╚═══════════════════════════════════════════════════════════════════════════════════╝
🎁🥚
During the development of Concave, we found a bug in the TiDB Vector Python client.
We fixed it using Concave Fleet!
pingcap/tidb-vector-python#58