Unique pr comment

.github/workflows/build-genai.yml

@@ -10,7 +10,9 @@ on:
             - "packages/core/**"
             - "packages/sample/**"
             - "packages/cli/**"
-
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     build:
         runs-on: ubuntu-latest

.github/workflows/build.yml

@@ -5,12 +5,15 @@ on:
         branches: [main]
     pull_request:
     workflow_dispatch:
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     build:
         runs-on: ubuntu-latest
         strategy:
-          matrix:
-              node-version: [20, 22]        
+            matrix:
+                node-version: [20, 22]
         steps:
             - uses: actions/checkout@v4
               with:

.github/workflows/genai-alt-text.yml

@@ -6,7 +6,9 @@ on:
     pull_request:
         paths:
             - "docs/src/**/*.png"
-
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     build:
         runs-on: ubuntu-latest

.github/workflows/genai-frontmatter.yml

@@ -7,6 +7,9 @@ on:
         paths:
             - "docs/src/**/*.md"
             - "docs/src/**/*.mdx"
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     build:
         runs-on: ubuntu-latest

.github/workflows/genai-pr-review.yml

@@ -1,13 +1,15 @@
 name: genai pull request review
 on:
     pull_request:
-        types: [ready_for_review]
         paths:
             - yarn.lock
             - ".github/workflows/ollama.yml"
             - "packages/core/**/*"
             - "packages/cli/**/*"
             - "packages/samples/**/*"
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
 jobs:
     build:
         runs-on: ubuntu-latest

.github/workflows/ollama.yml

@@ -17,6 +17,9 @@ on:
             - "packages/core/**/*"
             - "packages/cli/**/*"
             - "packages/samples/**/*"
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     tests:
         runs-on: ubuntu-latest

docs/src/content/docs/reference/cli/commands.md

@@ -22,10 +22,9 @@ Options:
   -od, --out-data <string>                   output file for data (.jsonl/ndjson will be aggregated). JSON schema information and validation will be included if available.
   -oa, --out-annotations <string>            output file for annotations (.csv will be rendered as csv, .jsonl/ndjson will be aggregated)
   -ocl, --out-changelog <string>             output file for changelogs
-  -prc, --pull-request-comment [string]      create comment on a pull request.
-  -prd, --pull-request-description [string]  upsert comment on a pull request description.
+  -prc, --pull-request-comment [string]      create comment on a pull request. use id to replace existing comment.
+  -prd, --pull-request-description [string]  create comment on a pull request description. Use id to replace existing comment.
   -prr, --pull-request-reviews               create pull request reviews from annotations
-  --no-pull-request-reviews-cache            disable pull request reviews cache
   -j, --json                                 emit full JSON response to output
   -y, --yaml                                 emit full YAML response to output
   -p, --prompt                               dry run, don't execute LLM and return expanded prompt

packages/cli/src/cli.ts

@@ -102,20 +102,16 @@ export async function cli() {
         .option("-ocl, --out-changelog <string>", "output file for changelogs")
         .option(
             "-prc, --pull-request-comment [string]",
-            "create comment on a pull request."
+            "create comment on a pull request with a unique id (defaults to script id)"
         )
         .option(
             "-prd, --pull-request-description [string]",
-            "upsert comment on a pull request description."
+            "create comment on a pull request description with a unique id (defaults to script id)"
         )
         .option(
             "-prr, --pull-request-reviews",
             "create pull request reviews from annotations"
         )
-        .option(
-            "--no-pull-request-reviews-cache",
-            "disable pull request reviews cache"
-        )
         .option("-j, --json", "emit full JSON response to output")
         .option("-y, --yaml", "emit full YAML response to output")
         .option(

packages/cli/src/run.ts

@@ -32,13 +32,9 @@ import {
     CSV_REGEX,
     CLI_RUN_FILES_FOLDER,
     parseGHTokenFromEnv,
-    githubUpsetPullRequest,
+    githubUpdatePullRequestDescription,
     githubCreatePullRequestReviews,
     githubCreateIssueComment,
-    GITHUB_PULL_REQUEST_REVIEWS_CACHE,
-    JSONLineCache,
-    PullRequestReviewsCacheKey,
-    PullRequestReviewsCacheValue,
 } from "genaiscript-core"
 import { capitalize } from "inflection"
 import { basename, resolve, join, relative } from "node:path"
@@ -66,7 +62,6 @@ export async function runScript(
         pullRequestComment: string | boolean
         pullRequestDescription: string | boolean
         pullRequestReviews: boolean
-        pullRequestReviewsCache: boolean
         outData: string
         label: string
         temperature: string
@@ -106,7 +101,6 @@ export async function runScript(
     const maxTokens = normalizeInt(options.maxTokens)
     const maxToolCalls = normalizeInt(options.maxToolCalls)
     const cache = !!options.cache
-    const pullRequestReviewsCache = !!options.pullRequestReviewsCache
     const applyEdits = !!options.applyEdits
     const csvSeparator = options.csvSeparator || "\t"
     const removeOut = options.removeOut
@@ -375,18 +369,7 @@ ${Array.from(files)
     if (pullRequestReviews && res.annotations?.length) {
         const info = parseGHTokenFromEnv(process.env)
         if (info.repository && info.issue) {
-            const cache = pullRequestReviewsCache
-                ? JSONLineCache.byName<
-                      PullRequestReviewsCacheKey,
-                      PullRequestReviewsCacheValue
-                  >(GITHUB_PULL_REQUEST_REVIEWS_CACHE)
-                : undefined
-            await githubCreatePullRequestReviews(
-                script,
-                info,
-                res.annotations,
-                { cache }
-            )
+            await githubCreatePullRequestReviews(script, info, res.annotations)
         }
     }
 
@@ -407,7 +390,7 @@ ${Array.from(files)
     if (pullRequestDescription && res.text) {
         const info = parseGHTokenFromEnv(process.env)
         if (info.repository && info.issue) {
-            await githubUpsetPullRequest(
+            await githubUpdatePullRequestDescription(
                 script,
                 info,
                 res.text,

packages/core/src/github.ts

@@ -1,15 +1,12 @@
 import { assert } from "node:console"
 import {
     GITHUB_API_VERSION,
-    GITHUB_PULL_REQUEST_REVIEWS_CACHE,
     GITHUB_TOKEN,
 } from "./constants"
 import { createFetch } from "./fetch"
 import { host } from "./host"
 import { link, prettifyMarkdown } from "./markdown"
 import { logError, logVerbose, normalizeInt } from "./util"
-import { string } from "mathjs"
-import { JSONLineCache } from "./cache"
 
 export interface GithubConnectionInfo {
     token: string
@@ -61,13 +58,14 @@ export function parseGHTokenFromEnv(
 }
 
 // https://docs.github.com/en/rest/pulls/pulls?apiVersion=2022-11-28#update-a-pull-request
-export async function githubUpsetPullRequest(
+export async function githubUpdatePullRequestDescription(
     script: PromptScript,
     info: GithubConnectionInfo,
     text: string,
     commentTag: string
 ) {
     const { apiUrl, repository, issue } = info
+    assert(commentTag)
 
     if (!issue) return { updated: false, statusText: "missing issue number" }
     const token = await host.readSecret(GITHUB_TOKEN)
@@ -89,15 +87,21 @@ export async function githubUpsetPullRequest(
     const resGetJson = (await resGet.json()) as { body: string }
     let { body } = resGetJson
     if (!body) body = ""
-    const tag = `\n\n<!-- genaiscript begin ${commentTag} -->\n\n`
-    const endTag = `\n\n<!-- genaiscript end ${commentTag} -->\n\n`
+    const tag = `<!-- genaiscript begin ${commentTag} -->`
+    const endTag = `<!-- genaiscript end ${commentTag} -->`
+    const sep = "\n\n"
 
     const start = body.indexOf(tag)
     const end = body.indexOf(endTag)
     if (start > -1 && end > -1 && start < end) {
-        body = body.slice(0, start + tag.length) + text + body.slice(end)
+        body =
+            body.slice(0, start + tag.length) +
+            sep +
+            text +
+            sep +
+            body.slice(end)
     } else {
-        body = body + tag + text + endTag
+        body = body + sep + tag + sep + text + sep + endTag + sep
     }
 
     const res = await fetch(url, {
@@ -213,20 +217,31 @@ async function githubCreatePullRequestReview(
     script: PromptScript,
     info: GithubConnectionInfo,
     token: string,
-    annotation: Diagnostic
+    annotation: Diagnostic,
+    existingComments: { id: string; path: string; line: number; body: string }[]
 ) {
     assert(token)
     const { apiUrl, repository, issue, commitSha } = info
 
-    const fetch = await createFetch({ retryOn: [] })
-    const url = `${apiUrl}/repos/${repository}/pulls/${issue}/comments`
     const body = {
         body: appendGeneratedComment(script, info, annotation.message),
         commit_id: commitSha,
         path: annotation.filename,
         line: annotation.range?.[0]?.[0],
         side: "RIGHT",
     }
+    if (
+        existingComments.find(
+            (c) =>
+                c.path === body.path &&
+                c.line === body.line &&
+                c.body === body.body
+        )
+    ) {
+        return { created: false, statusText: "comment already exists" }
+    }
+    const fetch = await createFetch({ retryOn: [] })
+    const url = `${apiUrl}/repos/${repository}/pulls/${issue}/comments`
     const res = await fetch(url, {
         method: "POST",
         headers: {
@@ -251,36 +266,12 @@ async function githubCreatePullRequestReview(
     return r
 }
 
-export interface PullRequestReviewsCacheKey {
-    repository: string
-    issue: number
-    scriptId: string
-    message: string
-    filename: string
-    line: number
-}
-
-export interface PullRequestReviewsCacheValue {
-    created: boolean
-    statusText: string
-    html_url: string
-}
-
-export type PullRequestReviewsCache = JSONLineCache<
-    PullRequestReviewsCacheKey,
-    PullRequestReviewsCacheValue
->
-
 export async function githubCreatePullRequestReviews(
     script: PromptScript,
     info: GithubConnectionInfo,
-    annotations: Diagnostic[],
-    options?: {
-        cache?: PullRequestReviewsCache
-    }
+    annotations: Diagnostic[]
 ): Promise<boolean> {
-    const { repository, issue, sha } = info
-    const { cache } = options || {}
+    const { repository, issue, sha, apiUrl } = info
 
     if (!annotations?.length) return true
     if (!issue) {
@@ -297,28 +288,33 @@ export async function githubCreatePullRequestReviews(
         return false
     }
 
+    // query existing reviews
+    const fetch = await createFetch({ retryOn: [] })
+    const url = `${apiUrl}/repos/${repository}/pulls/${issue}/comments`
+    const resListComments = await fetch(`${url}?per_page=100`, {
+        headers: {
+            Accept: "application/vnd.github+json",
+            Authorization: `Bearer ${token}`,
+            "X-GitHub-Api-Version": GITHUB_API_VERSION,
+        },
+    })
+    if (resListComments.status !== 200) return false
+    const comments = (await resListComments.json()) as {
+        id: string
+        path: string
+        line: number
+        body: string
+    }[]
+    console.log(comments)
     // code annotations
     for (const annotation of annotations) {
-        const cacheKey = {
-            repository,
-            issue,
-            scriptId: script.id,
-            message: annotation.message,
-            filename: annotation.filename,
-            line: annotation.range?.[0]?.[0],
-        }
-        const cached = await cache?.get(cacheKey)
-        if (cached)
-            logVerbose("ignore cached pull request review, " + cached.html_url)
-        else {
-            const res = await githubCreatePullRequestReview(
-                script,
-                info,
-                token,
-                annotation
-            )
-            if (res.created) await cache?.set(cacheKey, res)
-        }
+        await githubCreatePullRequestReview(
+            script,
+            info,
+            token,
+            annotation,
+            comments
+        )
     }
     return true
 }