Skip to content
View CoralGarden52's full-sized avatar
😉
😉
  • South China University of Technology

Block or report CoralGarden52

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
CoralGarden52/README.md

Open-source contributions

I contribute bug fixes, performance improvements, and reliability improvements to open-source AI, agent, and developer-tool projects.

Project PR What I fixed
Dify ★ #33116 Updated the backend contributor guide to recommend TypedDict for fixed-shape dictionary data.
Dify ★ #41400 Delete multimodal dataset attachment blobs after database cleanup, with tenant-scoped reads, safe storage deletion, and failure-tolerant logging.
Dify ★ #41999 Clean replaced multimodal attachments during reindexing and remove vectors/files only when they are no longer referenced.
Dify ★ #42159 Preserve CSV cell values as text during batch segment import, including leading zeros and literal NA values.
Dify ★ #42191 Replace per-document indexing-status count queries with grouped queries while preserving tenant filters and response semantics.
Dify ★ #42205 Enforce the intended ten-item cap for QA indexing previews by checking the active QA preview list.
Dify ★ #42233 Avoid one MAX(position) query per imported row by reading the current maximum once and incrementing it in memory.
MLflow ★ #25458 Remove the stale Helm image pin so the example follows Chart.appVersion by default.
MLflow ★ #25459 Move repository and ref inputs to actions/checkout so workflows check out the selected revision.
MLflow ★ #25474 Fix Docker examples that incorrectly expanded MLFLOW_BACKEND_STORE_URI in the host shell instead of using the container environment.
MLflow ★ #25477 Make the wheel-installation check test the exact commit checked out and used for the build.
MLflow ★ #25509 Preserve explicitly supplied empty webhook descriptions and secrets in partial updates using protobuf field presence.
MLflow ★ #25623 Normalize empty-list and NumPy-target evaluation inputs without IndexError or ambiguous array truth-value failures.
MLflow ★ #25768 Convert experiment_id from protobuf integer to string during REST hydration so online scoring updates use the required type.
Microsoft Agent Framework ★ #8116 Preserve ordered text, URI, and inline-data parts plus MIME-aware source fields in Python AG-UI chat requests.
Microsoft Agent Framework ★ #8224 Deduplicate MessagePack FileHistoryProvider writes consistently with the JSONL persistence path.
Microsoft Agent Framework ★ #8278 Count exceptions raised at the async await boundary against FunctionTool.max_invocation_exceptions.
Microsoft Agent Framework ★ #8300 Preserve functional-workflow user state across response-only human-in-the-loop resume.
Microsoft Agent Framework ★ #8309 Validate and coerce functional-workflow human-in-the-loop responses against the declared response type.
Microsoft Agent Framework ★ #8328 Resolve postponed response-handler annotations with get_type_hints while retaining a fallback for unresolved references.
Composio ★ #4297 Resolve internal $ref/$defs references before translating Vertex AI Google provider tool schemas.
Composio ★ #4298 Enforce authConfigId filtering so trigger subscriptions ignore events from other authentication configurations.
Composio ★ #4310 Preserve the original request order when merging mixed local and remote multi-execute results.
Composio ★ #4311 Qualify custom toolkit child-slug mappings across Python and TypeScript and fix related response mapping bugs.
Composio ★ #4448 Contain asynchronous Pusher subscription errors at the SDK boundary instead of allowing uncaught exceptions.
Haystack ★ #12501 Prevent FilterPolicy.MERGE from mutating stored filters and leaking runtime conditions across retriever runs.
Haystack ★ #12533 Clear stale metadata-extraction error state after a successful retry, including an empty JSON result.
Haystack ★ #12629 Allow an explicitly empty runtime filter dictionary to clear FilterRetriever defaults.
Haystack ★ #12639 Allow tools=[] to clear initialized tools for OpenAI chat generators while preserving defaults when omitted.
Haystack ★ #12748 Support composite JSON-serializable metadata in unique counts by normalizing nested lists, tuples, and dictionaries into hashable keys for consistent sync and async behavior.
AgentScope ★ #2519 Clear the inbox-consumer marker when deleting a session so reused IDs receive wakeups correctly.
AgentScope ★ #2521 Route channel hints through the inbox hand-off protocol during run finalization so queued messages retain their wakeups.
PEFT ★ #3667 Reject direct adapter injection for tuners with shared adapter state and direct users to get_peft_model with a clear error.
PEFT ★ #3714 Remove duplicated shared RandLoRA projections from adapter checkpoints while preserving backward-compatible loading, direct-injection support, and save_projection=False behavior.
Pydantic AI ★ #7541 Preserve dynamic native-factory configuration in XSearch and ImageGeneration fallback subagents and reject None results instead of enabling default tools.
FastMCP ★ #4948 Partition response-cache entries by component version so default and explicit-version requests cannot share stale results.
OpenAI Agents Python ★ #4780 Retry pre-request WebSocket handshake failures before any request frame is sent, while preserving no-replay behavior after transmission.

Pinned Loading

  1. mlflow/mlflow mlflow/mlflow Public

    The open source AI engineering platform for agents, LLMs, and ML models. MLflow enables teams of all sizes to debug, evaluate, monitor, and optimize production-quality AI applications while control…

    Python 28.2k 6.4k

  2. deepset-ai/haystack deepset-ai/haystack Public

    Open-source AI orchestration framework for building context-engineered, production-ready LLM applications. Design modular pipelines and agent workflows with explicit control over retrieval, routing…

    Python 26.6k 3.2k

  3. ComposioHQ/composio ComposioHQ/composio Public

    Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.

    TypeScript 30.4k 4.8k

  4. agentscope-ai/agentscope agentscope-ai/agentscope Public

    Build and run agents you can see, understand and trust.

    Python 32.6k 3.6k

  5. langgenius/dify langgenius/dify Public

    Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without …

    TypeScript 158k 24.9k

  6. microsoft/agent-framework microsoft/agent-framework Public

    A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET.

    Python 13.9k 2.4k