veryfront 0.1.99 → 0.1.100

package/esm/cli/commands/knowledge/command-help.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"command-help.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/command-help.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAEvD,eAAO,MAAM,aAAa,EAAE,WAuC3B,CAAC"}
+{"version":3,"file":"command-help.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/command-help.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAEvD,eAAO,MAAM,aAAa,EAAE,WAwC3B,CAAC"}

package/esm/cli/commands/knowledge/command-help.js

@@ -34,7 +34,8 @@ export const knowledgeHelp = {
         "Primary subcommand: ingest",
         "`uploads/...` means a remote project upload; use `./uploads/...` or `/workspace/uploads/...` to force a local file",
         "`ingest` orchestrates upload resolution, parsing, and project file writes",
-        "Requires python3; non-text formats also require the supported parser packages unless you run inside the Veryfront sandbox",
-        "The Veryfront sandbox image includes `kreuzberg`, and knowledge ingest falls back to the built-in parser when `kreuzberg` is unavailable or extraction fails",
+        "Requires python3; install `docling` locally to match the sandbox parsing path for PDF, Office, and HTML sources",
+        "Supported PDF, Office, and HTML sources are parsed through `docling` when it is available",
+        "The Veryfront sandbox image includes `docling`, and knowledge ingest falls back to the built-in parser when `docling` is unavailable or extraction fails",
     ],
 };

package/esm/cli/commands/knowledge/command.d.ts

@@ -17,6 +17,12 @@ export interface KnowledgeParserResult {
     stats: Record<string, unknown>;
     warnings: string[];
 }
+export interface KnowledgeParserInput {
+    filePath: string;
+    description?: string;
+    slug?: string;
+    sourceReference?: string;
+}
 type KnowledgeSource = {
     kind: "local";
     input: string;
@@ -128,6 +134,11 @@ export declare function runKnowledgeParser(input: {
     sourceReference?: string;
     env?: Record<string, string>;
 }): Promise<KnowledgeParserResult>;
+export declare function runKnowledgeParsers(input: {
+    files: KnowledgeParserInput[];
+    outputDir: string;
+    env?: Record<string, string>;
+}): Promise<KnowledgeParserResult[]>;
 export declare function collectKnowledgeSources(options: Pick<KnowledgeIngestOptions, "sources" | "path" | "all" | "recursive">, deps: {
     client: ApiClient;
     projectSlug: string;

package/esm/cli/commands/knowledge/command.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"command.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/command.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAKxB,OAAO,EAAE,KAAK,SAAS,EAA0C,MAAM,wBAAwB,CAAC;AAChG,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,uBAAuB,CAAC;AAIxD,OAAO,EAAuB,KAAK,MAAM,EAAgB,MAAM,6BAA6B,CAAC;AAG7F,OAAO,EAEL,KAAK,+BAA+B,EAEpC,KAAK,yBAAyB,EAC9B,KAAK,gCAAgC,EACtC,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,qBAAqB;IACpC,OAAO,EAAE,IAAI,CAAC;IACd,WAAW,EAAE,MAAM,CAAC;IACpB,eAAe,EAAE,MAAM,CAAC;IACxB,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE,MAAM,CAAC;IACb,mBAAmB,EAAE,MAAM,CAAC;IAC5B,sBAAsB,EAAE,MAAM,CAAC;IAC/B,WAAW,EAAE,MAAM,CAAC;IACpB,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC/B,QAAQ,EAAE,MAAM,EAAE,CAAC;CACpB;AAED,KAAK,eAAe,GAChB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;CAAE,GACnD;IAAE,IAAI,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;CAAE,CAAC;AAE7E,MAAM,WAAW,yBAAyB;IACxC,OAAO,EAAE,eAAe,EAAE,CAAC;IAC3B,OAAO,EAAE,gCAAgC,EAAE,CAAC;CAC7C;AAED,KAAK,cAAc,GAAG;IAAE,UAAU,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,CAAC;AAIhF,QAAA,MAAM,yBAAyB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAmD7B,CAAC;AAEH,MAAM,MAAM,sBAAsB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,yBAAyB,CAAC,CAAC;AAoD/E,wBAAgB,wBAAwB,CACtC,IAAI,EAAE,UAAU,GACf,CAAC,CAAC,mBAAmB,CAAC,OAAO,EAAE,sBAAsB,CAAC,CAexD;AAED,wBAAgB,2BAA2B,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAMrE;AAED,wBAAgB,0BAA0B,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAGpE;AAED,wBAAgB,2BAA2B,CAAC,UAAU,EAAE,MAAM,GAAG,MAAM,CAKtE;AAcD,wBAAgB,iBAAiB,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAGxD;AAyBD,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAE1D;AAMD,wBAAgB,iCAAiC,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAE3E;AAwID,wBAAgB,kBAAkB,CAAC,MAAM,EAAE,eAAe,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAkCjF;AACD,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,eAAe,EAAE,GAAG,MAAM,EAAE,CAQtE;AAED,wBAAgB,yBAAyB,CACvC,UAAU,EAAE,MAAM,EAClB,SAAS,EAAE,MAAM,EACjB,aAAa,EAAE,MAAM,GACpB,MAAM,CAQR;AAED,wBAAgB,2BAA2B,CAAC,KAAK,EAAE;IACjD,MAAM,EAAE,MAAM,CAAC;IACf,eAAe,EAAE,MAAM,CAAC;IACxB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,IAAI,CAAC,qBAAqB,EAAE,MAAM,GAAG,OAAO,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,CAAC,CAAC;CAChG,GAAG,yBAAyB,CAY5B;AAED,wBAAsB,kBAAkB,CAAC,KAAK,EAAE;IAC9C,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,GAAG,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CAC9B,GAAG,OAAO,CAAC,qBAAqB,CAAC,CAuDjC;AAED,wBAAsB,uBAAuB,CAC3C,OAAO,EAAE,IAAI,CAAC,sBAAsB,EAAE,SAAS,GAAG,MAAM,GAAG,KAAK,GAAG,WAAW,CAAC,EAC/E,IAAI,EAAE;IACJ,MAAM,EAAE,SAAS,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,eAAe,EAAE,CAAC,WAAW,EAAE,MAAM,EAAE,KAAK,OAAO,CAAC,cAAc,EAAE,CAAC,CAAC;CACvE,GACA,OAAO,CAAC,yBAAyB,CAAC,CAuJpC;AAED,wBAAsB,qBAAqB,CACzC,OAAO,EAAE,eAAe,EAAE,EAC1B,OAAO,EAAE,sBAAsB,EAC/B,IAAI,EAAE;IACJ,MAAM,EAAE,SAAS,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,EAAE,OAAO,kBAAkB,CAAC;IACrC,mBAAmB,EAAE,CAAC,UAAU,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,KAAK,OAAO,CAAC;QAAE,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IAC1F,WAAW,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;CAC7B,GACA,OAAO,CAAC;IACT,QAAQ,EAAE,yBAAyB,EAAE,CAAC;IACtC,MAAM,EAAE,+BAA+B,EAAE,CAAC;CAC3C,CAAC,CAsGD;AAED,wBAAsB,gBAAgB,CAAC,IAAI,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAwItE"}
+{"version":3,"file":"command.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/command.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAKxB,OAAO,EAAE,KAAK,SAAS,EAA0C,MAAM,wBAAwB,CAAC;AAChG,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,uBAAuB,CAAC;AAIxD,OAAO,EAAuB,KAAK,MAAM,EAAgB,MAAM,6BAA6B,CAAC;AAG7F,OAAO,EAEL,KAAK,+BAA+B,EAEpC,KAAK,yBAAyB,EAC9B,KAAK,gCAAgC,EACtC,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,qBAAqB;IACpC,OAAO,EAAE,IAAI,CAAC;IACd,WAAW,EAAE,MAAM,CAAC;IACpB,eAAe,EAAE,MAAM,CAAC;IACxB,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE,MAAM,CAAC;IACb,mBAAmB,EAAE,MAAM,CAAC;IAC5B,sBAAsB,EAAE,MAAM,CAAC;IAC/B,WAAW,EAAE,MAAM,CAAC;IACpB,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC/B,QAAQ,EAAE,MAAM,EAAE,CAAC;CACpB;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,eAAe,CAAC,EAAE,MAAM,CAAC;CAC1B;AACD,KAAK,eAAe,GAChB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;CAAE,GACnD;IAAE,IAAI,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;CAAE,CAAC;AAE7E,MAAM,WAAW,yBAAyB;IACxC,OAAO,EAAE,eAAe,EAAE,CAAC;IAC3B,OAAO,EAAE,gCAAgC,EAAE,CAAC;CAC7C;AAED,KAAK,cAAc,GAAG;IAAE,UAAU,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,CAAC;AAIhF,QAAA,MAAM,yBAAyB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAmD7B,CAAC;AAEH,MAAM,MAAM,sBAAsB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,yBAAyB,CAAC,CAAC;AAoD/E,wBAAgB,wBAAwB,CACtC,IAAI,EAAE,UAAU,GACf,CAAC,CAAC,mBAAmB,CAAC,OAAO,EAAE,sBAAsB,CAAC,CAexD;AAED,wBAAgB,2BAA2B,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAMrE;AAED,wBAAgB,0BAA0B,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAGpE;AAED,wBAAgB,2BAA2B,CAAC,UAAU,EAAE,MAAM,GAAG,MAAM,CAKtE;AAcD,wBAAgB,iBAAiB,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAGxD;AAyBD,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAE1D;AAMD,wBAAgB,iCAAiC,CAAC,SAAS,EAAE,MAAM,GAAG,MAAM,CAE3E;AAwID,wBAAgB,kBAAkB,CAAC,MAAM,EAAE,eAAe,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAkCjF;AACD,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,eAAe,EAAE,GAAG,MAAM,EAAE,CAQtE;AAED,wBAAgB,yBAAyB,CACvC,UAAU,EAAE,MAAM,EAClB,SAAS,EAAE,MAAM,EACjB,aAAa,EAAE,MAAM,GACpB,MAAM,CAQR;AAED,wBAAgB,2BAA2B,CAAC,KAAK,EAAE;IACjD,MAAM,EAAE,MAAM,CAAC;IACf,eAAe,EAAE,MAAM,CAAC;IACxB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,IAAI,CAAC,qBAAqB,EAAE,MAAM,GAAG,OAAO,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,CAAC,CAAC;CAChG,GAAG,yBAAyB,CAY5B;AAED,wBAAsB,kBAAkB,CAAC,KAAK,EAAE;IAC9C,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,GAAG,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CAC9B,GAAG,OAAO,CAAC,qBAAqB,CAAC,CAiBjC;AAED,wBAAsB,mBAAmB,CAAC,KAAK,EAAE;IAC/C,KAAK,EAAE,oBAAoB,EAAE,CAAC;IAC9B,SAAS,EAAE,MAAM,CAAC;IAClB,GAAG,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CAC9B,GAAG,OAAO,CAAC,qBAAqB,EAAE,CAAC,CA8DnC;AAED,wBAAsB,uBAAuB,CAC3C,OAAO,EAAE,IAAI,CAAC,sBAAsB,EAAE,SAAS,GAAG,MAAM,GAAG,KAAK,GAAG,WAAW,CAAC,EAC/E,IAAI,EAAE;IACJ,MAAM,EAAE,SAAS,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,eAAe,EAAE,CAAC,WAAW,EAAE,MAAM,EAAE,KAAK,OAAO,CAAC,cAAc,EAAE,CAAC,CAAC;CACvE,GACA,OAAO,CAAC,yBAAyB,CAAC,CAuJpC;AAED,wBAAsB,qBAAqB,CACzC,OAAO,EAAE,eAAe,EAAE,EAC1B,OAAO,EAAE,sBAAsB,EAC/B,IAAI,EAAE;IACJ,MAAM,EAAE,SAAS,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,EAAE,OAAO,kBAAkB,CAAC;IACrC,mBAAmB,EAAE,CAAC,UAAU,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,KAAK,OAAO,CAAC;QAAE,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IAC1F,WAAW,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;CAC7B,GACA,OAAO,CAAC;IACT,QAAQ,EAAE,yBAAyB,EAAE,CAAC;IACtC,MAAM,EAAE,+BAA+B,EAAE,CAAC;CAC3C,CAAC,CAsGD;AAED,wBAAsB,gBAAgB,CAAC,IAAI,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAwItE"}

package/esm/cli/commands/knowledge/command.js

@@ -348,6 +348,25 @@ export function createKnowledgeIngestResult(input) {
     };
 }
 export async function runKnowledgeParser(input) {
+    const [result] = await runKnowledgeParsers({
+        files: [{
+                filePath: input.filePath,
+                description: input.description,
+                slug: input.slug,
+                sourceReference: input.sourceReference,
+            }],
+        outputDir: input.outputDir,
+        env: input.env,
+    });
+    if (!result) {
+        throw new Error("knowledge ingest parser returned no results");
+    }
+    return result;
+}
+export async function runKnowledgeParsers(input) {
+    if (!input.files.length) {
+        return [];
+    }
     const tempDir = await dntShim.Deno.makeTempDir({ prefix: "veryfront-knowledge-parser-" });
     const inputJsonPath = `${tempDir}/input.json`;
     const outputJsonPath = `${tempDir}/output.json`;
@@ -355,11 +374,13 @@ export async function runKnowledgeParser(input) {
     try {
         try {
             await dntShim.Deno.writeTextFile(inputJsonPath, JSON.stringify({
-                file_path: input.filePath,
+                files: input.files.map((file) => ({
+                    file_path: file.filePath,
+                    description: file.description,
+                    slug: file.slug,
+                    source_reference: file.sourceReference,
+                })),
                 output_dir: input.outputDir,
-                description: input.description,
-                slug: input.slug,
-                source_reference: input.sourceReference,
             }));
             await dntShim.Deno.writeTextFile(scriptPath, knowledgeIngestPythonSource);
             let result;
@@ -382,7 +403,8 @@ export async function runKnowledgeParser(input) {
                 throw new Error(stderr || "parser exited unsuccessfully");
             }
             const raw = await dntShim.Deno.readTextFile(outputJsonPath);
-            return JSON.parse(raw);
+            const parsed = JSON.parse(raw);
+            return Array.isArray(parsed) ? parsed : [parsed];
         }
         catch (error) {
             if (error instanceof Error && error.message.startsWith("knowledge ingest parser failed")) {

package/esm/cli/commands/knowledge/parser-source.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"parser-source.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/parser-source.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,2BAA2B,QA2jBvC,CAAC"}
+{"version":3,"file":"parser-source.d.ts","sourceRoot":"","sources":["../../../../src/cli/commands/knowledge/parser-source.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,2BAA2B,QAgoBvC,CAAC"}

package/esm/cli/commands/knowledge/parser-source.js

@@ -2,8 +2,10 @@ export const knowledgeIngestPythonSource = String.raw `#!/usr/bin/env python3
 import argparse
 import csv
 import json
+import os
 import re
 import subprocess
+import tempfile
 from datetime import date
 from pathlib import Path
 from typing import Any, Optional
@@ -14,6 +16,26 @@ def yaml_quote(value: Any) -> str:
 
 
 CODE_FENCE = chr(96) * 3
+DEFAULT_DOCLING_TIMEOUT_SECONDS = 900.0
+
+
+def read_timeout_seconds(env_name: str, default_seconds: float) -> float:
+    raw_value = os.environ.get(env_name)
+    if raw_value is None or raw_value.strip() == "":
+        return default_seconds
+
+    try:
+        timeout_seconds = float(raw_value)
+    except ValueError:
+        return default_seconds
+
+    return timeout_seconds if timeout_seconds > 0 else default_seconds
+
+
+DOCLING_TIMEOUT_SECONDS = read_timeout_seconds(
+    "VERYFRONT_KNOWLEDGE_DOCLING_TIMEOUT_SECONDS",
+    DEFAULT_DOCLING_TIMEOUT_SECONDS,
+)
 TEXT_FILE_EXTENSIONS = {
     ".c",
     ".cc",
@@ -119,100 +141,68 @@ def build_frontmatter(source: str, source_type: str, description: str) -> str:
     ])
 
 
-def metadata_int(metadata: dict[str, Any], *keys: str) -> Optional[int]:
-    for key in keys:
-        value = metadata.get(key)
-        if isinstance(value, int) and not isinstance(value, bool):
-            return value
-    return None
-
-
-def metadata_string_list(metadata: dict[str, Any], *keys: str) -> Optional[list[str]]:
-    for key in keys:
-        value = metadata.get(key)
-        if isinstance(value, list) and all(isinstance(item, str) for item in value):
-            return value
-    return None
-
-
-def build_kreuzberg_stats(source_type: str, content: str, metadata: dict[str, Any]):
-    stats: dict[str, Any] = {
+def build_docling_stats(content: str):
+    return {
         "characters": len(content),
         "lines": len(content.splitlines()) if content else 0,
-        "engine": "kreuzberg",
+        "engine": "docling",
     }
 
-    if isinstance(metadata.get("mime_type"), str):
-        stats["mime_type"] = metadata["mime_type"]
 
-    if source_type == "pdf":
-        stats["pages"] = metadata_int(metadata, "page_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type in {"xlsx", "xls"}:
-        stats["sheets"] = metadata_int(metadata, "sheet_count") or 0
-        stats["rows"] = metadata_int(metadata, "row_count") or 0
-        stats["sheet_names"] = metadata_string_list(metadata, "sheet_names") or []
-    elif source_type == "docx":
-        stats["paragraphs"] = metadata_int(metadata, "paragraph_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type == "pptx":
-        stats["slides"] = metadata_int(metadata, "slide_count", "page_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type == "html":
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-
-    return stats
-
-
-def parse_with_kreuzberg(path: str, source_type: str):
-    warnings: list[str] = []
-    completed = subprocess.run(
-        [
-            "kreuzberg",
-            "extract",
-            path,
-            "--format",
-            "json",
-            "--output-format",
-            "markdown",
-        ],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-
-    if completed.returncode != 0:
-        detail = completed.stderr.strip() or completed.stdout.strip() or f"exit code {completed.returncode}"
-        raise RuntimeError(f"kreuzberg extract failed: {detail}")
+def run_docling_markdown(path: str):
+    with tempfile.TemporaryDirectory(prefix="veryfront-docling-") as output_dir:
+        try:
+            completed = subprocess.run(
+                [
+                    "docling",
+                    path,
+                    "--to",
+                    "md",
+                    "--image-export-mode",
+                    "placeholder",
+                    "--output",
+                    output_dir,
+                ],
+                capture_output=True,
+                text=True,
+                check=False,
+                timeout=DOCLING_TIMEOUT_SECONDS,
+            )
+        except subprocess.TimeoutExpired as error:
+            raise RuntimeError(
+                f"docling conversion timed out after {DOCLING_TIMEOUT_SECONDS:g}s"
+            ) from error
 
-    try:
-        payload = json.loads(completed.stdout)
-    except json.JSONDecodeError as error:
-        raise RuntimeError(f"kreuzberg extract returned invalid JSON: {error}") from error
+        if completed.returncode != 0:
+            detail = completed.stderr.strip() or completed.stdout.strip() or f"exit code {completed.returncode}"
+            raise RuntimeError(f"docling conversion failed: {detail}")
 
-    content = payload.get("content", "")
-    if not isinstance(content, str):
-        raise RuntimeError("kreuzberg extract did not return string content")
+        markdown_files = sorted(Path(output_dir).rglob("*.md"))
+        if not markdown_files:
+            raise RuntimeError("docling conversion did not produce a markdown file")
 
-    metadata = payload.get("metadata") if isinstance(payload.get("metadata"), dict) else {}
-    normalized_content = clean_text(content)
-    stats = build_kreuzberg_stats(source_type, normalized_content, metadata)
+        return markdown_files[0].read_text(encoding="utf-8")
 
+
+def parse_with_docling(path: str):
+    warnings: list[str] = []
+    normalized_content = clean_text(run_docling_markdown(path))
+    stats = build_docling_stats(normalized_content)
     return normalized_content or "_No extractable text found in document._", stats, warnings
 
 
-def prefer_kreuzberg(source_type: str, fallback_parser):
+def prefer_docling(fallback_parser):
     def parser(path: str):
         try:
-            return parse_with_kreuzberg(path, source_type)
+            return parse_with_docling(path)
         except FileNotFoundError as error:
-            if getattr(error, "filename", "") == "kreuzberg":
+            if getattr(error, "filename", "") == "docling":
                 return fallback_parser(path)
             raise
         except RuntimeError as error:
             content, stats, warnings = fallback_parser(path)
             warnings.append(
-                "kreuzberg extraction failed; fell back to the built-in parser: "
+                "docling conversion failed; fell back to the built-in parser: "
                 + str(error)
             )
             return content, stats, warnings
@@ -220,6 +210,10 @@ def prefer_kreuzberg(source_type: str, fallback_parser):
     return parser
 
 
+def build_parser(fallback_parser, prefers_docling: bool):
+    return prefer_docling(fallback_parser) if prefers_docling else fallback_parser
+
+
 def parse_csv_like(path: str, delimiter: str = ","):
     warnings: list[str] = []
     with open(path, newline="", encoding="utf-8-sig") as file:
@@ -451,36 +445,41 @@ def parse_json(path: str):
     return f"{CODE_FENCE}json\n{rendered}\n{CODE_FENCE}", stats, warnings
 
 
-def select_parser(path: Path):
+def select_parser_definition(path: Path):
     ext = path.suffix.lower()
     name = path.name.lower()
     if ext == ".pdf":
-        return "pdf", prefer_kreuzberg("pdf", parse_pdf)
+        return "pdf", parse_pdf, True
     if ext in {".csv", ".tsv"}:
         delimiter = "\t" if ext == ".tsv" else ","
-        return ext.lstrip("."), lambda file_path: parse_csv_like(file_path, delimiter)
+        return ext.lstrip("."), lambda file_path: parse_csv_like(file_path, delimiter), False
     if ext in {".xlsx", ".xls"}:
         source_type = ext.lstrip(".")
-        return source_type, prefer_kreuzberg(source_type, parse_excel)
+        return source_type, parse_excel, True
     if ext == ".docx":
-        return "docx", prefer_kreuzberg("docx", parse_docx)
+        return "docx", parse_docx, True
     if ext == ".pptx":
-        return "pptx", prefer_kreuzberg("pptx", parse_pptx)
+        return "pptx", parse_pptx, True
     if ext in {".html", ".htm"}:
-        return "html", prefer_kreuzberg("html", parse_html)
+        return "html", parse_html, True
     if ext in {".txt", ".md", ".mdx"}:
-        return ext.lstrip("."), parse_text
+        return ext.lstrip("."), parse_text, False
     if ext == ".json":
-        return "json", parse_json
+        return "json", parse_json, False
     if ext in TEXT_FILE_EXTENSIONS:
-        return ext.lstrip("."), parse_text
+        return ext.lstrip("."), parse_text, False
     if not ext and name in TEXT_FILE_NAMES:
-        return "text", parse_text
+        return "text", parse_text, False
     raise ValueError(f"Unsupported file type: {ext}")
 
 
+def select_parser(path: Path):
+    source_type, fallback_parser, prefers_docling = select_parser_definition(path)
+    return source_type, build_parser(fallback_parser, prefers_docling)
+
+
 def build_summary(source_type: str, stats: dict[str, Any]) -> str:
-    if stats.get("engine") == "kreuzberg":
+    if stats.get("engine") == "docling":
         return f"Converted {source_type.upper()} to markdown ({stats.get('characters', 0)} chars)."
     if source_type in {"csv", "tsv"}:
         return f"Parsed {stats.get('rows', 0)} rows across {stats.get('columns', 0)} columns."
@@ -540,6 +539,69 @@ def ingest_document_to_knowledge(file_path: str, output_dir: Optional[str] = Non
     }
 
 
+def ingest_documents_to_knowledge(documents: list[dict[str, Any]], output_dir: Optional[str] = None):
+    output_root = Path(output_dir or "/workspace/knowledge")
+    output_root.mkdir(parents=True, exist_ok=True)
+
+    prepared_documents: list[dict[str, Any]] = []
+    for index, document in enumerate(documents):
+        file_path = document["file_path"]
+        path = Path(file_path)
+        if not path.exists():
+            raise FileNotFoundError(f"File not found: {file_path}")
+
+        slug = document.get("slug") or slugify(path.stem)
+        source_type, fallback_parser, prefers_docling = select_parser_definition(path)
+        prepared_documents.append({
+            "index": index,
+            "path": path,
+            "slug": slug,
+            "description": document.get("description"),
+            "source_reference": document.get("source_reference"),
+            "source_type": source_type,
+            "fallback_parser": fallback_parser,
+            "prefers_docling": prefers_docling,
+        })
+
+    results = []
+    for document in prepared_documents:
+        parser = build_parser(
+            document["fallback_parser"],
+            document["prefers_docling"],
+        )
+        content, stats, warnings = parser(str(document["path"]))
+
+        content = clean_text(content)
+        resolved_description = document["description"] or f"Parsed from {document['path'].name}"
+        title = titleize_filename(document["path"])
+        frontmatter = build_frontmatter(
+            document["source_reference"] or document["path"].name,
+            document["source_type"],
+            resolved_description,
+        )
+        markdown = f"{frontmatter}\n\n# {title}\n\n{content}\n"
+
+        output_path = output_root / f"{document['slug']}.md"
+        output_path.write_text(markdown, encoding="utf-8")
+
+        results.append({
+            "success": True,
+            "source_path": str(document["path"]),
+            "source_filename": document["path"].name,
+            "source_type": document["source_type"],
+            "slug": document["slug"],
+            "sandbox_output_path": str(output_path),
+            "suggested_project_path": f"knowledge/{document['slug']}.md",
+            "description": resolved_description,
+            "title": title,
+            "summary": build_summary(document["source_type"], stats),
+            "stats": stats,
+            "warnings": warnings,
+        })
+
+    return results
+
+
 def main():
     parser = argparse.ArgumentParser(description="Convert a local document into knowledge-base markdown")
     parser.add_argument("--input-json", required=True)
@@ -548,20 +610,27 @@ def main():
 
     try:
         payload = json.loads(Path(args.input_json).read_text(encoding="utf-8"))
-        result = ingest_document_to_knowledge(
-            file_path=payload["file_path"],
-            output_dir=payload.get("output_dir"),
-            description=payload.get("description"),
-            slug=payload.get("slug"),
-            source_reference=payload.get("source_reference"),
-        )
+        files_payload = payload.get("files")
+        if isinstance(files_payload, list):
+            result = ingest_documents_to_knowledge(
+                documents=files_payload,
+                output_dir=payload.get("output_dir"),
+            )
+        else:
+            result = ingest_document_to_knowledge(
+                file_path=payload["file_path"],
+                output_dir=payload.get("output_dir"),
+                description=payload.get("description"),
+                slug=payload.get("slug"),
+                source_reference=payload.get("source_reference"),
+            )
     except ModuleNotFoundError as error:
         missing_package = error.name or "required package"
         raise SystemExit(
             "Missing Python package '"
             + missing_package
             + "'. Install knowledge parser dependencies with: "
-            + "pip install pandas openpyxl xlrd pdfplumber python-docx python-pptx beautifulsoup4 lxml"
+            + "pip install docling pandas openpyxl xlrd pdfplumber python-docx python-pptx beautifulsoup4 lxml"
         )
 
     Path(args.output_json).write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")

package/esm/deno.js

@@ -1,6 +1,6 @@
 export default {
     "name": "veryfront",
-    "version": "0.1.99",
+    "version": "0.1.100",
     "license": "Apache-2.0",
     "nodeModulesDir": "auto",
     "exclude": [
@@ -237,7 +237,7 @@ export default {
         "rehype-stringify": "npm:rehype-stringify@10.0.1",
         "esbuild": "npm:esbuild@0.27.4",
         "esbuild/mod.js": "npm:esbuild@0.27.4",
-        "es-module-lexer": "npm:es-module-lexer@1.5.0",
+        "es-module-lexer": "npm:es-module-lexer@2.0.0",
         "gray-matter": "npm:gray-matter@4.0.3",
         "zod": "npm:zod@3.25.76",
         "mime-types": "npm:mime-types@2.1.35",

package/esm/src/utils/version.d.ts

@@ -1,4 +1,4 @@
-export declare const VERSION = "0.1.99";
+export declare const VERSION = "0.1.100";
 export declare function normalizeVeryfrontVersion(version: string | undefined): string | undefined;
 export declare function resolveRuntimeVersion(options?: {
     veryfrontVersion?: string;

package/esm/src/utils/version.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"version.d.ts","sourceRoot":"","sources":["../../../src/src/utils/version.ts"],"names":[],"mappings":"AAKA,eAAO,MAAM,OAAO,WAAW,CAAC;AAEhC,wBAAgB,yBAAyB,CAAC,OAAO,EAAE,MAAM,GAAG,SAAS,GAAG,MAAM,GAAG,SAAS,CAGzF;AAUD,wBAAgB,qBAAqB,CAAC,OAAO,GAAE;IAC7C,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,eAAe,CAAC,EAAE,MAAM,CAAC;CACrB,GAAG,MAAM,CAKd;AAED,eAAO,MAAM,eAAe,QAK1B,CAAC;AAEH,eAAO,MAAM,iBAAiB,EAAE,MAAmB,CAAC;AAEpD,MAAM,WAAW,YAAY;IAC3B,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,wBAAgB,kBAAkB,CAAC,gBAAgB,CAAC,EAAE,MAAM,GAAG,YAAY,CAM1E"}
+{"version":3,"file":"version.d.ts","sourceRoot":"","sources":["../../../src/src/utils/version.ts"],"names":[],"mappings":"AAKA,eAAO,MAAM,OAAO,YAAY,CAAC;AAEjC,wBAAgB,yBAAyB,CAAC,OAAO,EAAE,MAAM,GAAG,SAAS,GAAG,MAAM,GAAG,SAAS,CAGzF;AAUD,wBAAgB,qBAAqB,CAAC,OAAO,GAAE;IAC7C,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,eAAe,CAAC,EAAE,MAAM,CAAC;CACrB,GAAG,MAAM,CAKd;AAED,eAAO,MAAM,eAAe,QAK1B,CAAC;AAEH,eAAO,MAAM,iBAAiB,EAAE,MAAmB,CAAC;AAEpD,MAAM,WAAW,YAAY;IAC3B,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;IACpB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,wBAAgB,kBAAkB,CAAC,gBAAgB,CAAC,EAAE,MAAM,GAAG,YAAY,CAM1E"}

package/esm/src/utils/version.js

@@ -2,7 +2,7 @@ import denoConfig from "../../deno.js";
 import { getEnv } from "../platform/compat/process.js";
 // Keep in sync with deno.json version.
 // scripts/release.ts updates this constant during releases.
-export const VERSION = "0.1.99";
+export const VERSION = "0.1.100";
 export function normalizeVeryfrontVersion(version) {
     if (!version)
         return undefined;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "veryfront",
-  "version": "0.1.99",
+  "version": "0.1.100",
   "description": "The simplest way to build AI-powered apps",
   "keywords": [
     "react",
@@ -148,7 +148,7 @@
     "ai": "6.0.134",
     "class-variance-authority": "0.7.1",
     "clsx": "2.1.1",
-    "es-module-lexer": "1.5.0",
+    "es-module-lexer": "2.0.0",
     "esbuild": "0.27.4",
     "github-slugger": "2.0.0",
     "gray-matter": "4.0.3",

package/src/cli/commands/knowledge/command-help.ts

@@ -36,7 +36,8 @@ export const knowledgeHelp: CommandHelp = {
     "Primary subcommand: ingest",
     "`uploads/...` means a remote project upload; use `./uploads/...` or `/workspace/uploads/...` to force a local file",
     "`ingest` orchestrates upload resolution, parsing, and project file writes",
-    "Requires python3; non-text formats also require the supported parser packages unless you run inside the Veryfront sandbox",
-    "The Veryfront sandbox image includes `kreuzberg`, and knowledge ingest falls back to the built-in parser when `kreuzberg` is unavailable or extraction fails",
+    "Requires python3; install `docling` locally to match the sandbox parsing path for PDF, Office, and HTML sources",
+    "Supported PDF, Office, and HTML sources are parsed through `docling` when it is available",
+    "The Veryfront sandbox image includes `docling`, and knowledge ingest falls back to the built-in parser when `docling` is unavailable or extraction fails",
   ],
 };

package/src/cli/commands/knowledge/command.ts

@@ -35,6 +35,12 @@ export interface KnowledgeParserResult {
   warnings: string[];
 }
 
+export interface KnowledgeParserInput {
+  filePath: string;
+  description?: string;
+  slug?: string;
+  sourceReference?: string;
+}
 type KnowledgeSource =
   | { kind: "local"; input: string; localPath: string }
   | { kind: "upload"; input: string; uploadPath: string; localPath: string };
@@ -465,6 +471,33 @@ export async function runKnowledgeParser(input: {
   sourceReference?: string;
   env?: Record<string, string>;
 }): Promise<KnowledgeParserResult> {
+  const [result] = await runKnowledgeParsers({
+    files: [{
+      filePath: input.filePath,
+      description: input.description,
+      slug: input.slug,
+      sourceReference: input.sourceReference,
+    }],
+    outputDir: input.outputDir,
+    env: input.env,
+  });
+
+  if (!result) {
+    throw new Error("knowledge ingest parser returned no results");
+  }
+
+  return result;
+}
+
+export async function runKnowledgeParsers(input: {
+  files: KnowledgeParserInput[];
+  outputDir: string;
+  env?: Record<string, string>;
+}): Promise<KnowledgeParserResult[]> {
+  if (!input.files.length) {
+    return [];
+  }
+
   const tempDir = await dntShim.Deno.makeTempDir({ prefix: "veryfront-knowledge-parser-" });
   const inputJsonPath = `${tempDir}/input.json`;
   const outputJsonPath = `${tempDir}/output.json`;
@@ -475,11 +508,13 @@ export async function runKnowledgeParser(input: {
       await dntShim.Deno.writeTextFile(
         inputJsonPath,
         JSON.stringify({
-          file_path: input.filePath,
+          files: input.files.map((file) => ({
+            file_path: file.filePath,
+            description: file.description,
+            slug: file.slug,
+            source_reference: file.sourceReference,
+          })),
           output_dir: input.outputDir,
-          description: input.description,
-          slug: input.slug,
-          source_reference: input.sourceReference,
         }),
       );
       await dntShim.Deno.writeTextFile(scriptPath, knowledgeIngestPythonSource);
@@ -507,7 +542,8 @@ export async function runKnowledgeParser(input: {
       }
 
       const raw = await dntShim.Deno.readTextFile(outputJsonPath);
-      return JSON.parse(raw) as KnowledgeParserResult;
+      const parsed = JSON.parse(raw) as KnowledgeParserResult | KnowledgeParserResult[];
+      return Array.isArray(parsed) ? parsed : [parsed];
     } catch (error) {
       if (error instanceof Error && error.message.startsWith("knowledge ingest parser failed")) {
         throw error;

package/src/cli/commands/knowledge/parser-source.ts

@@ -2,8 +2,10 @@ export const knowledgeIngestPythonSource = String.raw`#!/usr/bin/env python3
 import argparse
 import csv
 import json
+import os
 import re
 import subprocess
+import tempfile
 from datetime import date
 from pathlib import Path
 from typing import Any, Optional
@@ -14,6 +16,26 @@ def yaml_quote(value: Any) -> str:
 
 
 CODE_FENCE = chr(96) * 3
+DEFAULT_DOCLING_TIMEOUT_SECONDS = 900.0
+
+
+def read_timeout_seconds(env_name: str, default_seconds: float) -> float:
+    raw_value = os.environ.get(env_name)
+    if raw_value is None or raw_value.strip() == "":
+        return default_seconds
+
+    try:
+        timeout_seconds = float(raw_value)
+    except ValueError:
+        return default_seconds
+
+    return timeout_seconds if timeout_seconds > 0 else default_seconds
+
+
+DOCLING_TIMEOUT_SECONDS = read_timeout_seconds(
+    "VERYFRONT_KNOWLEDGE_DOCLING_TIMEOUT_SECONDS",
+    DEFAULT_DOCLING_TIMEOUT_SECONDS,
+)
 TEXT_FILE_EXTENSIONS = {
     ".c",
     ".cc",
@@ -119,100 +141,68 @@ def build_frontmatter(source: str, source_type: str, description: str) -> str:
     ])
 
 
-def metadata_int(metadata: dict[str, Any], *keys: str) -> Optional[int]:
-    for key in keys:
-        value = metadata.get(key)
-        if isinstance(value, int) and not isinstance(value, bool):
-            return value
-    return None
-
-
-def metadata_string_list(metadata: dict[str, Any], *keys: str) -> Optional[list[str]]:
-    for key in keys:
-        value = metadata.get(key)
-        if isinstance(value, list) and all(isinstance(item, str) for item in value):
-            return value
-    return None
-
-
-def build_kreuzberg_stats(source_type: str, content: str, metadata: dict[str, Any]):
-    stats: dict[str, Any] = {
+def build_docling_stats(content: str):
+    return {
         "characters": len(content),
         "lines": len(content.splitlines()) if content else 0,
-        "engine": "kreuzberg",
+        "engine": "docling",
     }
 
-    if isinstance(metadata.get("mime_type"), str):
-        stats["mime_type"] = metadata["mime_type"]
 
-    if source_type == "pdf":
-        stats["pages"] = metadata_int(metadata, "page_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type in {"xlsx", "xls"}:
-        stats["sheets"] = metadata_int(metadata, "sheet_count") or 0
-        stats["rows"] = metadata_int(metadata, "row_count") or 0
-        stats["sheet_names"] = metadata_string_list(metadata, "sheet_names") or []
-    elif source_type == "docx":
-        stats["paragraphs"] = metadata_int(metadata, "paragraph_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type == "pptx":
-        stats["slides"] = metadata_int(metadata, "slide_count", "page_count") or 0
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-    elif source_type == "html":
-        stats["tables"] = metadata_int(metadata, "table_count") or 0
-
-    return stats
-
-
-def parse_with_kreuzberg(path: str, source_type: str):
-    warnings: list[str] = []
-    completed = subprocess.run(
-        [
-            "kreuzberg",
-            "extract",
-            path,
-            "--format",
-            "json",
-            "--output-format",
-            "markdown",
-        ],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-
-    if completed.returncode != 0:
-        detail = completed.stderr.strip() or completed.stdout.strip() or f"exit code {completed.returncode}"
-        raise RuntimeError(f"kreuzberg extract failed: {detail}")
+def run_docling_markdown(path: str):
+    with tempfile.TemporaryDirectory(prefix="veryfront-docling-") as output_dir:
+        try:
+            completed = subprocess.run(
+                [
+                    "docling",
+                    path,
+                    "--to",
+                    "md",
+                    "--image-export-mode",
+                    "placeholder",
+                    "--output",
+                    output_dir,
+                ],
+                capture_output=True,
+                text=True,
+                check=False,
+                timeout=DOCLING_TIMEOUT_SECONDS,
+            )
+        except subprocess.TimeoutExpired as error:
+            raise RuntimeError(
+                f"docling conversion timed out after {DOCLING_TIMEOUT_SECONDS:g}s"
+            ) from error
 
-    try:
-        payload = json.loads(completed.stdout)
-    except json.JSONDecodeError as error:
-        raise RuntimeError(f"kreuzberg extract returned invalid JSON: {error}") from error
+        if completed.returncode != 0:
+            detail = completed.stderr.strip() or completed.stdout.strip() or f"exit code {completed.returncode}"
+            raise RuntimeError(f"docling conversion failed: {detail}")
 
-    content = payload.get("content", "")
-    if not isinstance(content, str):
-        raise RuntimeError("kreuzberg extract did not return string content")
+        markdown_files = sorted(Path(output_dir).rglob("*.md"))
+        if not markdown_files:
+            raise RuntimeError("docling conversion did not produce a markdown file")
 
-    metadata = payload.get("metadata") if isinstance(payload.get("metadata"), dict) else {}
-    normalized_content = clean_text(content)
-    stats = build_kreuzberg_stats(source_type, normalized_content, metadata)
+        return markdown_files[0].read_text(encoding="utf-8")
 
+
+def parse_with_docling(path: str):
+    warnings: list[str] = []
+    normalized_content = clean_text(run_docling_markdown(path))
+    stats = build_docling_stats(normalized_content)
     return normalized_content or "_No extractable text found in document._", stats, warnings
 
 
-def prefer_kreuzberg(source_type: str, fallback_parser):
+def prefer_docling(fallback_parser):
     def parser(path: str):
         try:
-            return parse_with_kreuzberg(path, source_type)
+            return parse_with_docling(path)
         except FileNotFoundError as error:
-            if getattr(error, "filename", "") == "kreuzberg":
+            if getattr(error, "filename", "") == "docling":
                 return fallback_parser(path)
             raise
         except RuntimeError as error:
             content, stats, warnings = fallback_parser(path)
             warnings.append(
-                "kreuzberg extraction failed; fell back to the built-in parser: "
+                "docling conversion failed; fell back to the built-in parser: "
                 + str(error)
             )
             return content, stats, warnings
@@ -220,6 +210,10 @@ def prefer_kreuzberg(source_type: str, fallback_parser):
     return parser
 
 
+def build_parser(fallback_parser, prefers_docling: bool):
+    return prefer_docling(fallback_parser) if prefers_docling else fallback_parser
+
+
 def parse_csv_like(path: str, delimiter: str = ","):
     warnings: list[str] = []
     with open(path, newline="", encoding="utf-8-sig") as file:
@@ -451,36 +445,41 @@ def parse_json(path: str):
     return f"{CODE_FENCE}json\n{rendered}\n{CODE_FENCE}", stats, warnings
 
 
-def select_parser(path: Path):
+def select_parser_definition(path: Path):
     ext = path.suffix.lower()
     name = path.name.lower()
     if ext == ".pdf":
-        return "pdf", prefer_kreuzberg("pdf", parse_pdf)
+        return "pdf", parse_pdf, True
     if ext in {".csv", ".tsv"}:
         delimiter = "\t" if ext == ".tsv" else ","
-        return ext.lstrip("."), lambda file_path: parse_csv_like(file_path, delimiter)
+        return ext.lstrip("."), lambda file_path: parse_csv_like(file_path, delimiter), False
     if ext in {".xlsx", ".xls"}:
         source_type = ext.lstrip(".")
-        return source_type, prefer_kreuzberg(source_type, parse_excel)
+        return source_type, parse_excel, True
     if ext == ".docx":
-        return "docx", prefer_kreuzberg("docx", parse_docx)
+        return "docx", parse_docx, True
     if ext == ".pptx":
-        return "pptx", prefer_kreuzberg("pptx", parse_pptx)
+        return "pptx", parse_pptx, True
     if ext in {".html", ".htm"}:
-        return "html", prefer_kreuzberg("html", parse_html)
+        return "html", parse_html, True
     if ext in {".txt", ".md", ".mdx"}:
-        return ext.lstrip("."), parse_text
+        return ext.lstrip("."), parse_text, False
     if ext == ".json":
-        return "json", parse_json
+        return "json", parse_json, False
     if ext in TEXT_FILE_EXTENSIONS:
-        return ext.lstrip("."), parse_text
+        return ext.lstrip("."), parse_text, False
     if not ext and name in TEXT_FILE_NAMES:
-        return "text", parse_text
+        return "text", parse_text, False
     raise ValueError(f"Unsupported file type: {ext}")
 
 
+def select_parser(path: Path):
+    source_type, fallback_parser, prefers_docling = select_parser_definition(path)
+    return source_type, build_parser(fallback_parser, prefers_docling)
+
+
 def build_summary(source_type: str, stats: dict[str, Any]) -> str:
-    if stats.get("engine") == "kreuzberg":
+    if stats.get("engine") == "docling":
         return f"Converted {source_type.upper()} to markdown ({stats.get('characters', 0)} chars)."
     if source_type in {"csv", "tsv"}:
         return f"Parsed {stats.get('rows', 0)} rows across {stats.get('columns', 0)} columns."
@@ -540,6 +539,69 @@ def ingest_document_to_knowledge(file_path: str, output_dir: Optional[str] = Non
     }
 
 
+def ingest_documents_to_knowledge(documents: list[dict[str, Any]], output_dir: Optional[str] = None):
+    output_root = Path(output_dir or "/workspace/knowledge")
+    output_root.mkdir(parents=True, exist_ok=True)
+
+    prepared_documents: list[dict[str, Any]] = []
+    for index, document in enumerate(documents):
+        file_path = document["file_path"]
+        path = Path(file_path)
+        if not path.exists():
+            raise FileNotFoundError(f"File not found: {file_path}")
+
+        slug = document.get("slug") or slugify(path.stem)
+        source_type, fallback_parser, prefers_docling = select_parser_definition(path)
+        prepared_documents.append({
+            "index": index,
+            "path": path,
+            "slug": slug,
+            "description": document.get("description"),
+            "source_reference": document.get("source_reference"),
+            "source_type": source_type,
+            "fallback_parser": fallback_parser,
+            "prefers_docling": prefers_docling,
+        })
+
+    results = []
+    for document in prepared_documents:
+        parser = build_parser(
+            document["fallback_parser"],
+            document["prefers_docling"],
+        )
+        content, stats, warnings = parser(str(document["path"]))
+
+        content = clean_text(content)
+        resolved_description = document["description"] or f"Parsed from {document['path'].name}"
+        title = titleize_filename(document["path"])
+        frontmatter = build_frontmatter(
+            document["source_reference"] or document["path"].name,
+            document["source_type"],
+            resolved_description,
+        )
+        markdown = f"{frontmatter}\n\n# {title}\n\n{content}\n"
+
+        output_path = output_root / f"{document['slug']}.md"
+        output_path.write_text(markdown, encoding="utf-8")
+
+        results.append({
+            "success": True,
+            "source_path": str(document["path"]),
+            "source_filename": document["path"].name,
+            "source_type": document["source_type"],
+            "slug": document["slug"],
+            "sandbox_output_path": str(output_path),
+            "suggested_project_path": f"knowledge/{document['slug']}.md",
+            "description": resolved_description,
+            "title": title,
+            "summary": build_summary(document["source_type"], stats),
+            "stats": stats,
+            "warnings": warnings,
+        })
+
+    return results
+
+
 def main():
     parser = argparse.ArgumentParser(description="Convert a local document into knowledge-base markdown")
     parser.add_argument("--input-json", required=True)
@@ -548,20 +610,27 @@ def main():
 
     try:
         payload = json.loads(Path(args.input_json).read_text(encoding="utf-8"))
-        result = ingest_document_to_knowledge(
-            file_path=payload["file_path"],
-            output_dir=payload.get("output_dir"),
-            description=payload.get("description"),
-            slug=payload.get("slug"),
-            source_reference=payload.get("source_reference"),
-        )
+        files_payload = payload.get("files")
+        if isinstance(files_payload, list):
+            result = ingest_documents_to_knowledge(
+                documents=files_payload,
+                output_dir=payload.get("output_dir"),
+            )
+        else:
+            result = ingest_document_to_knowledge(
+                file_path=payload["file_path"],
+                output_dir=payload.get("output_dir"),
+                description=payload.get("description"),
+                slug=payload.get("slug"),
+                source_reference=payload.get("source_reference"),
+            )
     except ModuleNotFoundError as error:
         missing_package = error.name or "required package"
         raise SystemExit(
             "Missing Python package '"
             + missing_package
             + "'. Install knowledge parser dependencies with: "
-            + "pip install pandas openpyxl xlrd pdfplumber python-docx python-pptx beautifulsoup4 lxml"
+            + "pip install docling pandas openpyxl xlrd pdfplumber python-docx python-pptx beautifulsoup4 lxml"
         )
 
     Path(args.output_json).write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")

package/src/deno.js

@@ -1,6 +1,6 @@
 export default {
   "name": "veryfront",
-  "version": "0.1.99",
+  "version": "0.1.100",
   "license": "Apache-2.0",
   "nodeModulesDir": "auto",
   "exclude": [
@@ -237,7 +237,7 @@ export default {
     "rehype-stringify": "npm:rehype-stringify@10.0.1",
     "esbuild": "npm:esbuild@0.27.4",
     "esbuild/mod.js": "npm:esbuild@0.27.4",
-    "es-module-lexer": "npm:es-module-lexer@1.5.0",
+    "es-module-lexer": "npm:es-module-lexer@2.0.0",
     "gray-matter": "npm:gray-matter@4.0.3",
     "zod": "npm:zod@3.25.76",
     "mime-types": "npm:mime-types@2.1.35",

package/src/src/transforms/esm/lexer.ts

@@ -60,7 +60,7 @@ export type ImportSpecifier = {
   ss: number; // Start of import statement
   se: number; // End of import statement
   d: number; // > -1 if dynamic import
-  a: number; // assert index
+  a: number; // import attribute index
 };
 
 function logParseError(error: unknown, code: string): void {

package/src/src/utils/version.ts

@@ -3,7 +3,7 @@ import { getEnv } from "../platform/compat/process.js";
 
 // Keep in sync with deno.json version.
 // scripts/release.ts updates this constant during releases.
-export const VERSION = "0.1.99";
+export const VERSION = "0.1.100";
 
 export function normalizeVeryfrontVersion(version: string | undefined): string | undefined {
   if (!version) return undefined;