Add functionality to diff GitHub Actions job logs and update related documentation and tool definitions

docs/src/components/BuiltinTools.mdx

@@ -20,7 +20,8 @@
 <LinkCard title="github_actions_workflows_list" description="List all workflows as a list of 'id: name' pair." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
 <LinkCard title="github_actions_runs_list" description="List all runs for a workflow. Use 'git_actions_list_workflows' to list workflows." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
 <LinkCard title="github_actions_jobs_list" description="List all jobs for a run." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
-<LinkCard title="github_actions_job_log" description="Download job log." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
+<LinkCard title="github_actions_job_logs_get" description="Download workflow job log." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
+<LinkCard title="github_actions_job_los_diff" description="Diffs two workflow job logs." href="/genaiscript/reference/scripts/system#systemgithub_actions" />
 <LinkCard title="github_files_get" description="Get a file from a repository." href="/genaiscript/reference/scripts/system#systemgithub_files" />
 <LinkCard title="github_files_list" description="List all files in a repository." href="/genaiscript/reference/scripts/system#systemgithub_files" />
 <LinkCard title="github_issues_list" description="List all issues in a repository." href="/genaiscript/reference/scripts/system#systemgithub_issues" />

docs/src/content/docs/reference/scripts/system.mdx

@@ -119,6 +119,7 @@
             type: "string",
             description: "Query to answer",
         },
+        required: ["query"],
     },
     async (args) => {
         const { context, query } = args
@@ -146,6 +147,7 @@
                     "system.retrieval_fuzz_search",
                     "system.md_frontmatter",
                 ],
+                label: "agent file system"
             }
         )
         return res
@@ -178,6 +180,7 @@
             type: "string",
             description: "Query to answer",
         },
+        required: ["query"],
     },
     async (args) => {
         const { context, query } = args
@@ -202,6 +205,7 @@
                     "system.explanations",
                     "system.git",
                 ],
+                label: "agent git"
             }
         )
         return res
@@ -234,6 +238,7 @@
             type: "string",
             description: "Query to answer",
         },
+        required: ["query"],
     },
     async (args) => {
         const { context, query } = args
@@ -246,8 +251,12 @@
 
                 Analyze and answer QUERY.
 
+                ## Constraints
+
                 - Assume that your answer will be analyzed by an LLM, not a human.
                 - If you cannot answer the query, return an empty string.
+                - Prefer diffing job logs rather downloading entire logs which can be very large.
+
                 `
             },
             {
@@ -261,6 +270,7 @@
                     "system.github_issues",
                     "system.github_pulls",
                 ],
+                label: "agent github",
             }
         )
         return res
@@ -796,7 +806,8 @@
 -  tool `github_actions_workflows_list`: List all workflows as a list of 'id: name' pair.
 -  tool `github_actions_runs_list`: List all runs for a workflow. Use 'git_actions_list_workflows' to list workflows.
 -  tool `github_actions_jobs_list`: List all jobs for a run.
--  tool `github_actions_job_log`: Download job log.
+-  tool `github_actions_job_logs_get`: Download workflow job log.
+-  tool `github_actions_job_los_diff`: Diffs two workflow job logs.
 
 `````js wrap title="system.github_actions"
 system({
@@ -881,8 +892,8 @@
 )
 
 defTool(
-    "github_actions_job_log",
-    "Download job log.",
+    "github_actions_job_logs_get",
+    "Download workflow job log.",
     {
         job_id: {
             type: "string",
@@ -900,6 +911,28 @@
     }
 )
 
+defTool(
+    "github_actions_job_los_diff",
+    "Diffs two workflow job logs.",
+    {
+        job_id: {
+            type: "string",
+            description: "ID of the job to compare.",
+        },
+        other_job_id: {
+            type: "string",
+            description: "ID of the other job to compare.",
+        },
+        required: ["job_id", "other_job_id"],
+    },
+    async (args) => {
+        const { job_id, other_job_id, context } = args
+        context.log(`github action diff job logs ${job_id} ${other_job_id}`)
+        const log = await github.diffWorkflowJobLogs(job_id, other_job_id)
+        return log
+    }
+)
+
 `````
 
 

packages/core/src/chat.ts

@@ -447,7 +447,7 @@
         cancellationToken,
     } = options
 
-    stats.addUsage(req.model, resp.usage)
+    stats.addUsage(req, resp.usage)
 
     if (resp.text)
         messages.push({

packages/core/src/genaisrc/system.agent_fs.genai.mjs

@@ -40,6 +40,7 @@ defTool(
                     "system.retrieval_fuzz_search",
                     "system.md_frontmatter",
                 ],
+                label: "agent file system"
             }
         )
         return res

packages/core/src/genaisrc/system.agent_git.genai.mjs

@@ -37,6 +37,7 @@ defTool(
                     "system.explanations",
                     "system.git",
                 ],
+                label: "agent git"
             }
         )
         return res

packages/core/src/genaisrc/system.agent_github.genai.mjs

@@ -25,8 +25,12 @@ defTool(
 
                 Analyze and answer QUERY.
 
+                ## Constraints
+
                 - Assume that your answer will be analyzed by an LLM, not a human.
                 - If you cannot answer the query, return an empty string.
+                - Prefer diffing job logs rather downloading entire logs which can be very large.
+
                 `
             },
             {
@@ -40,6 +44,7 @@ defTool(
                     "system.github_issues",
                     "system.github_pulls",
                 ],
+                label: "agent github",
             }
         )
         return res

packages/core/src/genaisrc/system.github_actions.genai.mjs

@@ -80,8 +80,8 @@ defTool(
 )
 
 defTool(
-    "github_actions_job_log",
-    "Download job log.",
+    "github_actions_job_logs_get",
+    "Download workflow job log.",
     {
         job_id: {
             type: "string",
@@ -98,3 +98,25 @@ defTool(
         return log
     }
 )
+
+defTool(
+    "github_actions_job_los_diff",
+    "Diffs two workflow job logs.",
+    {
+        job_id: {
+            type: "string",
+            description: "ID of the job to compare.",
+        },
+        other_job_id: {
+            type: "string",
+            description: "ID of the other job to compare.",
+        },
+        required: ["job_id", "other_job_id"],
+    },
+    async (args) => {
+        const { job_id, other_job_id, context } = args
+        context.log(`github action diff job logs ${job_id} ${other_job_id}`)
+        const log = await github.diffWorkflowJobLogs(job_id, other_job_id)
+        return log
+    }
+)

packages/core/src/git.ts

@@ -82,10 +82,11 @@ export class GitClient implements Git {
         }
     ): Promise<WorkspaceFile[]> {
         const { askStageOnEmpty } = options || {}
-        const paths = arrayify(options?.paths).filter((f) => !!f)
-        const excludedPaths = arrayify(options?.excludedPaths).filter(
-            (f) => !!f
-        )
+        const paths = arrayify(options?.paths, { filterEmpty: true })
+        const excludedPaths = arrayify(options?.excludedPaths, {
+            filterEmpty: true,
+        })
+
         let filenames: string[]
         if (scope === "modified-base" || scope === "staged") {
             const args = ["diff", "--name-only", "--diff-filter=AM"]
@@ -177,10 +178,10 @@ export class GitClient implements Git {
         excludedPaths?: ElementOrArray<string>
     }): Promise<GitCommit[]> {
         const { base, head, merges, excludedGrep } = options || {}
-        const paths = arrayify(options?.paths).filter((f) => !!f)
-        const excludedPaths = arrayify(options?.excludedPaths).filter(
-            (f) => !!f
-        )
+        const paths = arrayify(options?.paths, { filterEmpty: true })
+        const excludedPaths = arrayify(options?.excludedPaths, {
+            filterEmpty: true,
+        })
 
         const args = ["log", "--pretty=oneline"]
         if (!merges) args.push("--no-merges")
@@ -192,7 +193,7 @@ export class GitClient implements Git {
             args.push(`--grep='${pattern}'`, "--invert-grep")
         }
         if (base && head) args.push(`${base}..${head}`)
-        GitClient.addFileFilters(arrayify(paths), arrayify(excludedPaths), args)
+        GitClient.addFileFilters(paths, excludedPaths, args)
         const res = await this.exec(args)
         const commits = res
             .split("\n")
@@ -221,10 +222,10 @@ export class GitClient implements Git {
         unified?: number
         llmify?: boolean
     }): Promise<string> {
-        const paths = arrayify(options?.paths).filter((f) => !!f)
-        const excludedPaths = arrayify(options?.excludedPaths).filter(
-            (f) => !!f
-        )
+        const paths = arrayify(options?.paths, { filterEmpty: true })
+        const excludedPaths = arrayify(options?.excludedPaths, {
+            filterEmpty: true,
+        })
         const { staged, base, head, unified, askStageOnEmpty } = options || {}
         const args = ["diff"]
         if (staged) args.push("--staged")

packages/core/src/github.ts

@@ -16,6 +16,7 @@
 import { isGlobMatch } from "./glob"
 import { fetchText } from "./fetch"
 import { concurrentLimit } from "./concurrency"
+import { createDiff, llmifyDiff } from "./diff"
 
 export interface GithubConnectionInfo {
     token: string
@@ -706,6 +707,31 @@
         return text
     }
 
+    private async downladJob(job_id: number) {
+        const { client, owner, repo } = await this.client()
+        const filename = `job-${job_id}.log`
+        const { url } = await client.rest.actions.downloadJobLogsForWorkflowRun(
+            {
+                owner,
+                repo,
+                job_id,
+            }
+        )
+        const { text: content } = await fetchText(url)
+        return { filename, url, content }
+    }
+
+    async diffWorkflowJobLogs(job_id: number, other_job_id: number) {
+        const job = await this.downladJob(job_id)
+        const other = await this.downladJob(other_job_id)
+
+        job.content = parseJobLog(job.content)
+        other.content = parseJobLog(other.content)
+
+        const diff = createDiff(job, other)
+        return llmifyDiff(diff)
+    }
+
     async getFile(filename: string, ref: string): Promise<WorkspaceFile> {
         const { client, owner, repo } = await this.client()
         const { data: content } = await client.rest.repos.getContent({

packages/core/src/types/prompt_template.d.ts

@@ -1531,6 +1531,11 @@ interface GitHub {
         options?: { llmify?: boolean }
     ): Promise<string>
 
+    /**
+     * Diffs two GitHub Action workflow job logs
+     */
+    diffWorkflowJobLogs(job_id: number, other_job_id: number): Promise<string>
+
     /**
      * Lists issues for a given repository
      * @param options