Deploy this function on flows.network, and you will get a GitHub agent to review and summarize Pull Requests. It helps busy open source contributors understand and make decisions on PRs faster! A few examples below!
We recommend you to use a GaiaNet node running an open source coding LLM as the backend to perform PR reviews and summarizations. You can use a community node or run a node on your own computer!
- [Rust] Improve support for host functions in the WasmEdge Rust SDK
- [bash] Support ARM architecture in the WasmEdge installer
- [C++] Add an eBPF plugin for WasmEdge
- [Haskell] Improve the CLI utility for WasmEdge Component Model tooling
Still not convinced? See "potential problems 1" in this review, it identified an inefficient Rust implementation of an algorithm. 🤯
This bot summarizes commits in the PR. Alternatively, you can use this bot to review changed files in the PR.
This flow function will be triggered when a new PR is raised in the designated GitHub repo. The flow function collects the content in the PR, and asks ChatGPT/4 to review and summarize it. The result is then posted back to the PR as a comment. The flow functions are written in Rust and run in hosted WasmEdge Runtimes on flows.network.
- The PR summary comment is updated automatically every time a new commit is pushed to this PR.
- A new summary could be triggered when someone says a magic trigger phrase in the PR's comments section. The default trigger phrase is "flows summarize".
- Create a bot from your own GitHub repo
- Configure the bot to review PRs
- Connect to GitHub for access to the target repo
You will also need to sign into flows.network from your GitHub account. It is free.
Just fork this repo to your own GitHub account.
Then, from flows.network, you can "Create a Flow" and select your forked repo. It will create a flow function based on the code in your forked repo.
Click on the "Advanced" button to see configuration options for the flow function.
Next, you will tell the bot which GitHub repo it needs to monitor for upcoming PRs to review.
github_owner: GitHub org for the repo you want to review PRsgithub_repo: GitHub repo you want to review PRstrigger_phrase: The magic words to write in a PR comment to summon the bot. It defaults to "flows summarize".
Let's see an example. You would like to deploy the bot to review code in PRs on
WasmEdge/wasmedge_hyper_demorepo. Heregithub_owner = WasmEdgeandgithub_repo = wasmedge_hyper_demo.
And the LLM API service you want to use to review the PRs.
llm_api_endpoint: The OpenAI compatible API service endpoint for the LLM to conduct code reviews. We recommend the Codetral-0.1-22B GaiaNet node:https://codestral-01-22b.us.gaianet.network/v1llm_model_name: The model name required by the API service. We recommend the following model name for the above GaiaNet node:Codestral-22B-v0.1-hf-Q5_K_Mllm_ctx_size: The context window size of the selected model. The Codestral 0.1 model has a 16k context window, which is16384.llm_api_key: Optional: The API key if required by the LLM service provider. It is not required for the GaiaNet node.
Click on the Build button.
Finally, the GitHub repo will need to give you access so that the flow function can access and review its PRs! In this next screen, you will connect to GitHub and authorize access.
Click on the Connect or + Add new authentication button to give the function access to the GitHub repo. You'll be redirected to a new page where you must grant flows.network permission to the repo.
Close the tab and go back to the flow.network page once you are done. Click on Deploy.
This is it! You are now on the flow details page waiting for the flow function to build. As soon as the flow's status became running, the bot is ready to give code reviews! The bot is summoned by every new PR, every new commit, as well as magic words (i.e., trigger_phrase) in PR comments.
You can manually create a new flow and import the source code repo for the bot (i.e., the repo you cloned from the template). Then, you can use the flow config to specify the github_owner and github_repo to point to the target repo you need to deploy the bot on. Deploy and authorize access to that target repo.
You can repeat this for all target repos you would like to deploy this bot on.
You could have a single flow function repo deployed as the source code for multiple bots. When you update the source code in the repo, and push it to GitHub, it will change the behavior of all the bots.
Go to the "Settings" tab of the running flow function for the bot, you can update the trigger_phrase config. The value of this config is the magic phrase the user will say to trigger a review from a PR comment.
This flow function is originally created by Jay Chen, and jinser made significant contributions to optimize the event triggers from GitHub.
