feat: Skip large diffs when generating context for the LLM

packages/cli/src/lib/processPatchset.ts

@@ -0,0 +1,24 @@
+/**
+ * Examine all the diffs in a git diff output and replace long ones with a simple
+ * message indicating that there is a change and its size.
+ * @param patchset - The patchset to process.
+ * @param maxDiffLength - The maximum length of a diff to include in the patchset.
+ * @returns The processed patchset.
+ */
+
+export default function processPatchset(patchset: string, maxDiffLength = 5000): string {
+  const parts = patchset.split(/^(?=diff|commit)/m);
+  return parts
+    .map((part) => {
+      if (part.startsWith('diff')) {
+        const lines = part.split('\n');
+        const header = lines[0];
+        const body = lines.slice(1).join('\n');
+        if (body.length > maxDiffLength) {
+          return `${header}\n[Change of size ${body.trimEnd().length}]\n`;
+        }
+      }
+      return part;
+    })
+    .join('');
+}

packages/cli/src/rpc/explain/review.ts

@@ -1,6 +1,7 @@
 import { parseOptions, REVIEW_DIFF_LOCATION, UserContext } from '@appland/navie';
 import configuration from '../configuration';
 import { getDiffLog, getWorkingDiff } from '../../lib/git';
+import processPatchset from '../../lib/processPatchset';
 
 /**
  * This function is responsible for transforming user context to include diff content when the
@@ -27,7 +28,10 @@ export default async function handleReview(
       {
         type: 'code-snippet',
         location: REVIEW_DIFF_LOCATION, // eslint-disable-line @typescript-eslint/no-unsafe-assignment
-        content: diffContent.filter(Boolean).join('\n\n'),
+        content: diffContent
+          .filter(Boolean)
+          .map((diff) => processPatchset(diff))
+          .join('\n\n'),
       },
     ],
   };

packages/cli/src/rpc/navie/welcome-suggestion.ts

@@ -1,5 +1,6 @@
 import configuration from '../configuration';
 import { getDiffLog, getWorkingDiff } from '../../lib/git';
+import processPatchset from '../../lib/processPatchset';
 import INavie, { INavieProvider } from '../explain/navie/inavie';
 import { UserContext } from '@appland/navie';
 import isCustomWelcomeMessageEnabled from './isCustomWelcomeMessageEnabled';
@@ -40,7 +41,7 @@ async function getChangeDiffs(projectDirectories: string[]): Promise<string[]> {
   );
   return diffs
     .filter((result): result is PromiseFulfilledResult<string> => result.status === 'fulfilled')
-    .map((result) => result.value);
+    .map((result) => processPatchset(result.value));
 }
 
 export async function getWelcomeSuggestion(

packages/cli/tests/unit/lib/processPatchset.spec.ts

@@ -0,0 +1,161 @@
+import processPatchset from '../../../src/lib/processPatchset';
+
+describe('processPatchset', () => {
+  it('should handle patchset with small diffs correctly', () => {
+    const patchset = `diff --git a/file1.txt b/file1.txt
+index 83db48f..f735c4e 100644
+--- a/file1.txt
++++ b/file1.txt
+@@ -1,3 +1,3 @@
+-Hello World
++Hello AppMap
+ This is a test file.`;
+
+    const result = processPatchset(patchset, 1000);
+    expect(result).toContain('Hello AppMap');
+  });
+
+  it('should have the correct output format', () => {
+    const largeDiff = 'a'.repeat(2000);
+    const patchset = `Here's some header.
+diff --git a/file1.txt b/file1.txt
+index 83db48f..f735c4e 100644
+--- a/file1.txt
++++ b/file1.txt
+@@ -1,3 +1,3 @@
+-Hello World
++Hello AppMap
+This is a test file.
+diff --git a/file2.txt b/file2.txt
+index 83db48f..f735c4e 100644
+--- a/file2.txt
++++ b/file2.txt
+@@ -1,3 +1,3 @@
+${largeDiff}
+diff --git a/file3.txt b/file3.txt
+index 83db48f..f735c4e 100644
+--- a/file3.txt
++++ b/file3.txt
+@@ -1,3 +1,3 @@
+-Hello World
++Hello AppMap
+ This is another test file.`;
+
+    const result = processPatchset(patchset, 1000);
+    expect(result).toMatchInlineSnapshot(`
+        "Here's some header.
+        diff --git a/file1.txt b/file1.txt
+        index 83db48f..f735c4e 100644
+        --- a/file1.txt
+        +++ b/file1.txt
+        @@ -1,3 +1,3 @@
+        -Hello World
+        +Hello AppMap
+        This is a test file.
+        diff --git a/file2.txt b/file2.txt
+        [Change of size 2078]
+        diff --git a/file3.txt b/file3.txt
+        index 83db48f..f735c4e 100644
+        --- a/file3.txt
+        +++ b/file3.txt
+        @@ -1,3 +1,3 @@
+        -Hello World
+        +Hello AppMap
+         This is another test file."
+      `);
+  });
+
+  it('should handle patchsets with multiple commit headers like from git log', () => {
+    const patchset = `commit 1
+Author: John Doe <[email protected]>
+Date:   Mon Jan 1 12:00:00 2023 +0000
+
+    Initial commit
+
+diff --git a/file1.txt b/file1.txt
+index 83db48f..f735c4e 100644
+--- a/file1.txt
++++ b/file1.txt
+@@ -1,3 +1,3 @@
+-Hello World
++Hello AppMap
+ This is a test file.
+
+commit 2
+Author: Jane Doe <[email protected]>
+Date:   Tue Jan 2 12:00:00 2023 +0000
+
+    Second commit
+
+diff --git a/file2.txt b/file2.txt
+index 83db48f..f735c4e 100644
+--- a/file2.txt
++++ b/file2.txt
+@@ -1,3 +1,3 @@
+-Hello World
++Hello AppMap
+ This is another test file.`;
+
+    const result = processPatchset(patchset, 10);
+    expect(result).toMatchInlineSnapshot(`
+        "commit 1
+        Author: John Doe <[email protected]>
+        Date:   Mon Jan 1 12:00:00 2023 +0000
+
+            Initial commit
+
+        diff --git a/file1.txt b/file1.txt
+        [Change of size 126]
+        commit 2
+        Author: Jane Doe <[email protected]>
+        Date:   Tue Jan 2 12:00:00 2023 +0000
+
+            Second commit
+
+        diff --git a/file2.txt b/file2.txt
+        [Change of size 132]
+        "
+      `);
+  });
+
+  it('should handle patchset with large diffs correctly', () => {
+    const largeDiff = 'a'.repeat(2000);
+    const patchset = `diff --git a/file2.txt b/file2.txt
+index 83db48f..f735c4e 100644
+--- a/file2.txt
++++ b/file2.txt
+@@ -1,3 +1,3 @@
+${largeDiff}`;
+
+    const result = processPatchset(patchset, 1000);
+    expect(result).toContain('[Change of size 2078]');
+  });
+
+  it('should handle patchset with mixed diff sizes correctly', () => {
+    const largeDiff = 'a'.repeat(2000);
+    const smallDiff = 'b'.repeat(500);
+    const patchset = `diff --git a/file3.txt b/file3.txt
+index 83db48f..f735c4e 100644
+--- a/file3.txt
++++ b/file3.txt
+@@ -1,3 +1,3 @@
+${largeDiff}
+diff --git a/file4.txt b/file4.txt
+index 83db48f..f735c4e 100644
+--- a/file4.txt
++++ b/file4.txt
+@@ -1,3 +1,3 @@
+${smallDiff}`;
+
+    const result = processPatchset(patchset, 1000);
+    expect(result).toContain('[Change of size 2078]');
+    expect(result).toContain(smallDiff);
+  });
+
+  it('should return other text unchanged', () => {
+    const diffs = ['This is some text with no diffs.', '', '- this\n+ that'];
+    const result = diffs.map((diff) => processPatchset(diff));
+    expect(result).toEqual(diffs);
+  });
+});
+