From d8b5633442f3a489ed937c71ade8145417583496 Mon Sep 17 00:00:00 2001
From: Jean-Baptiste THERY <jb.thery@jcode.works>
Date: Mon, 24 Aug 2026 14:42:21 +0700
Subject: [PATCH] fix(core): support multi-document YAML ingestion

Release highlights:
- Index multi-document YAML streams without rejecting valid project files.

Release details:
- Parse YAML streams in order, omit null documents, and preserve parser errors.
- Cover multi-document, single-document, empty-document, and invalid-stream regressions.

Verification:
- pnpm validate
- Isolated rgr ingest and search reproduction for issue #159.

Fixes #159
---
 packages/ragmir-core/src/parsing.test.ts | 50 ++++++++++++++++++++++++
 packages/ragmir-core/src/parsing.ts      | 14 ++++++-
 2 files changed, 63 insertions(+), 1 deletion(-)

diff --git a/packages/ragmir-core/src/parsing.test.ts b/packages/ragmir-core/src/parsing.test.ts
index 2dbfb0b..9177a45 100644
--- a/packages/ragmir-core/src/parsing.test.ts
+++ b/packages/ragmir-core/src/parsing.test.ts
@@ -38,6 +38,56 @@ describe("parseFile", () => {
     expect(parsed.text).not.toContain("hidden image")
   })
 
+  it("should concatenate non-empty YAML documents in source order", async () => {
+    const root = await mkdtemp(path.join(os.tmpdir(), "ragmir-yaml-stream-"))
+    tempDirs.push(root)
+    const filePath = path.join(root, "context.yaml")
+    await writeFile(
+      filePath,
+      [
+        "---",
+        'project_context: "test"',
+        "quality_score: 5",
+        "---",
+        "---",
+        "# Title",
+        "",
+        "This is the body of the YAML document.",
+      ].join("\n"),
+      "utf8",
+    )
+
+    const parsed = await parseFile(sourceFile(root, filePath, ".yaml"))
+
+    expect(parsed.text).toContain("project_context")
+    expect(parsed.text).toContain("quality_score: 5")
+    expect(parsed.text).toContain("This is the body of the YAML document.")
+    expect(parsed.text.indexOf("project_context")).toBeLessThan(
+      parsed.text.indexOf("This is the body of the YAML document."),
+    )
+    expect(parsed.text).not.toMatch(/(^|\n)null($|\n)/u)
+  })
+
+  it("should preserve single-document YAML parsing", async () => {
+    const root = await mkdtemp(path.join(os.tmpdir(), "ragmir-yaml-single-"))
+    tempDirs.push(root)
+    const filePath = path.join(root, "context.yml")
+    await writeFile(filePath, "project: ragmir\nquality: 5\n", "utf8")
+
+    const parsed = await parseFile(sourceFile(root, filePath, ".yml"))
+
+    expect(parsed.text).toBe("project: ragmir\nquality: 5")
+  })
+
+  it("should reject invalid documents in a YAML stream", async () => {
+    const root = await mkdtemp(path.join(os.tmpdir(), "ragmir-yaml-invalid-"))
+    tempDirs.push(root)
+    const filePath = path.join(root, "invalid.yaml")
+    await writeFile(filePath, "---\nvalid: true\n---\nbroken: [\n", "utf8")
+
+    await expect(parseFile(sourceFile(root, filePath, ".yaml"))).rejects.toThrow(/Flow sequence/iu)
+  })
+
   it("extracts text from docx files", async () => {
     const root = await mkdtemp(path.join(os.tmpdir(), "ragmir-docx-"))
     tempDirs.push(root)
diff --git a/packages/ragmir-core/src/parsing.ts b/packages/ragmir-core/src/parsing.ts
index e0d65b0..2be2cc2 100644
--- a/packages/ragmir-core/src/parsing.ts
+++ b/packages/ragmir-core/src/parsing.ts
@@ -118,7 +118,19 @@ export async function parseFile(
     }
     case ".yaml":
     case ".yml": {
-      text = YAML.stringify(YAML.parse(await readFile(file.absolutePath, "utf8")))
+      const documents = YAML.parseAllDocuments(await readFile(file.absolutePath, "utf8"))
+      for (const document of documents) {
+        const [error] = document.errors
+        if (error) {
+          throw error
+        }
+      }
+      text = documents
+        .flatMap((document) => {
+          const value = document.toJS()
+          return value === null ? [] : [YAML.stringify(value)]
+        })
+        .join("\n")
       break
     }
     case ".rtf":