Fix Pixelbot post-merge validation issues

.agents/skills/pixeltable-skill/SKILL.md → .agents/skills/pixeltable/SKILL.md

@@ -12,7 +12,7 @@ license: Apache-2.0
 allowed-tools: []
 metadata:
   author: Pixeltable
-  version: 2.10.0
+  version: 2.10.1
   type: documentation
   executes-code: false
   category: data-infrastructure

.agents/skills/pixeltable-skill/references/core-api.md → .agents/skills/pixeltable/references/core-api.md

@@ -248,7 +248,7 @@ The cell is set to `None`, nothing raises, and `errormsg` stays empty. This is b
 ```python
 @pxt.udf
 def label(severity: str | None) -> str:
-    return payload or 'unknown'
+    return severity or 'unknown'
 ```
 
 Binding a nullable argument to a non-nullable parameter also widens the column's declared type, so `excerpt(Docs.body)` over `body: pxt.String | None` is a `String | None` column.

.github/workflows/ci.yml

@@ -13,8 +13,8 @@ jobs:
       matrix:
         python-version: ["3.11", "3.12", "3.13", "3.14"]
     steps:
-      - uses: actions/checkout@v7
-      - uses: astral-sh/setup-uv@v10
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: astral-sh/setup-uv@bec219d24cd3e171d82865faccec33120bb574f4 # v10.1.0
         with:
           python-version: ${{ matrix.python-version }}
       - run: uv sync --locked --group dev
@@ -32,8 +32,8 @@ jobs:
   frontend:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v7
-      - uses: actions/setup-node@v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
         with:
           node-version: "22.12"
           cache: npm
@@ -50,8 +50,8 @@ jobs:
   package:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v7
-      - uses: astral-sh/setup-uv@v10
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: astral-sh/setup-uv@bec219d24cd3e171d82865faccec33120bb574f4 # v10.1.0
         with:
           python-version: "3.11"
       - run: uv sync --locked --group dev
@@ -67,8 +67,8 @@ jobs:
   pixeltable-schema:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v7
-      - uses: astral-sh/setup-uv@v10
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: astral-sh/setup-uv@bec219d24cd3e171d82865faccec33120bb574f4 # v10.1.0
         with:
           python-version: "3.11"
       - run: uv sync --locked --group dev

AGENTS.md

@@ -9,7 +9,9 @@ Pixelbot 3.0 is an app-first Pixeltable 0.7.7 project. Python 3.11+ and Node 22.
 - `backend/pixelbot/routers/`: custom REST routes. The database router is read-only.
 - `backend/pixelbot/static/`: production SPA package data generated by Vite.
 - `frontend/`: React SPA with route-level lazy loading.
-- `.agents/skills/pixeltable-skill/`: bundled current Pixeltable guidance.
+- `.agents/skills/pixeltable/`: a reproducible project-local snapshot of the canonical
+  `pixeltable/pixeltable-skill` repository. `skills-lock.json` records its source and content hash;
+  keep both in sync when refreshing the skill.
 
 ## Catalog rules
 

backend/pixelbot/app.py

@@ -89,6 +89,15 @@ def user_info() -> dict[str, str]:
     return {"user_name": config.DEFAULT_USER_NAME}
 
 
+@app.api_route(
+    "/api/{unmatched_path:path}",
+    methods=["GET", "POST", "PUT", "PATCH", "DELETE"],
+    include_in_schema=False,
+)
+async def unmatched_api_route(unmatched_path: str) -> JSONResponse:
+    return JSONResponse(status_code=404, content={"detail": "Not Found"})
+
+
 staticDir = Path(__file__).resolve().parent / "static"
 
 if staticDir.is_dir():

backend/pixelbot/routers/database.py

@@ -444,8 +444,8 @@ def join_tables(body: JoinRequest):
     "search_documents": "pixelbot_v3/chunks",
     "search_images": "pixelbot_v3/images",
     "search_video_frames": "pixelbot_v3/video_frames",
-    "search_video_transcripts": "pixelbot_v3/video_transcript_sentences",
-    "search_audio_transcripts": "pixelbot_v3/audio_transcript_sentences",
+    "search_video_transcripts": "pixelbot_v3/video_audio_chunks",
+    "search_audio_transcripts": "pixelbot_v3/audio_chunks",
     "search_memory": "pixelbot_v3/memory_bank",
     "search_chat_history": "pixelbot_v3/chat_history",
     "get_recent_chat_history": "pixelbot_v3/chat_history",

backend/pixelbot/routers/studio.py

@@ -1008,7 +1008,7 @@ def search_studio(body: SearchRequest):
 
         # Also search video transcripts
         try:
-            vt_view = pxt.get_table("pixelbot_v3.video_transcript_sentences")
+            vt_view = pxt.get_table("pixelbot_v3.video_audio_chunks")
             sim = vt_view.text.similarity(string=body.query)
             for row in (
                 vt_view.where((vt_view.user_id == user_id) & (sim > body.threshold))
@@ -1031,7 +1031,7 @@ def search_studio(body: SearchRequest):
     # ── Audio transcripts (Gemini embed) ──
     if "audio" in body.types:
         try:
-            at_view = pxt.get_table("pixelbot_v3.audio_transcript_sentences")
+            at_view = pxt.get_table("pixelbot_v3.audio_chunks")
             sim = at_view.text.similarity(string=body.query)
             for row in (
                 at_view.where((at_view.user_id == user_id) & (sim > body.threshold))
@@ -1190,9 +1190,9 @@ def _collect_text_embeddings(
     except Exception as e:
         logger.error(f"Embedding viz: document error: {e}")
 
-    # Video transcript sentences
+    # Video transcript chunks
     try:
-        vt_view = pxt.get_table("pixelbot_v3.video_transcript_sentences")
+        vt_view = pxt.get_table("pixelbot_v3.video_audio_chunks")
         for row in (
             vt_view.where(vt_view.user_id == user_id)
             .select(
@@ -1216,9 +1216,9 @@ def _collect_text_embeddings(
     except Exception as e:
         logger.error(f"Embedding viz: video transcript error: {e}")
 
-    # Audio transcript sentences
+    # Audio transcript chunks
     try:
-        at_view = pxt.get_table("pixelbot_v3.audio_transcript_sentences")
+        at_view = pxt.get_table("pixelbot_v3.audio_chunks")
         for row in (
             at_view.where(at_view.user_id == user_id)
             .select(
@@ -2280,9 +2280,9 @@ def get_transcription(uuid: str, media_type: str):
 
     try:
         if media_type == "audio":
-            view_name = "pixelbot_v3.audio_transcript_sentences"
+            view_name = "pixelbot_v3.audio_chunks"
         else:
-            view_name = "pixelbot_v3.video_transcript_sentences"
+            view_name = "pixelbot_v3.video_audio_chunks"
 
         view = pxt.get_table(view_name)
         sentences = []

backend/pixelbot/schema.py

@@ -6,6 +6,8 @@
 
 from __future__ import annotations
 
+from typing import Any
+
 import pixeltable as pxt
 from pixeltable.functions import bfl, gemini, openai
 from pixeltable.functions import image as pxt_image
@@ -14,7 +16,6 @@
 from pixeltable.functions.document import document_splitter
 from pixeltable.functions.gemini import invoke_tools
 from pixeltable.functions.huggingface import clip
-from pixeltable.functions.string import string_splitter
 from pixeltable.functions.video import extract_audio, frame_iterator
 
 from pixelbot import config, functions
@@ -26,6 +27,31 @@
 clipEmbed = clip.using(model_id=config.CLIP_MODEL_ID)
 
 
+def _jsonb_stable(value: Any) -> Any:
+    """Match PostgreSQL JSONB key ordering for stable schema reconciliation."""
+    if isinstance(value, dict):
+        return {key: _jsonb_stable(value[key]) for key in sorted(value, key=lambda key: (len(key), key))}
+    if isinstance(value, list):
+        return [_jsonb_stable(item) for item in value]
+    return value
+
+
+DOCUMENT_SUMMARY_CONFIG = _jsonb_stable(
+    {
+        "system_instruction": "Analyze the document text and return a structured summary.",
+        "response_mime_type": "application/json",
+        "response_schema": DocumentSummary.model_json_schema(),
+    }
+)
+FOLLOW_UP_CONFIG = _jsonb_stable(
+    {
+        "system_instruction": "Generate exactly 3 relevant follow-up questions based on the conversation.",
+        "response_mime_type": "application/json",
+        "response_schema": FollowUpResponse.model_json_schema(),
+    }
+)
+
+
 class Documents(TableModel, name="collection", has_default_idxs=False):
     document: pxt.Document
     uuid = pxt.Column(type=pxt.String, primary_key=True)
@@ -35,11 +61,7 @@ class Documents(TableModel, name="collection", has_default_idxs=False):
     summary_response = gemini.generate_content(
         contents=document_text,
         model=config.GEMINI_MODEL_ID,
-        config={
-            "system_instruction": "Analyze the document text and return a structured summary.",
-            "response_mime_type": "application/json",
-            "response_schema": DocumentSummary.model_json_schema(),
-        },
+        config=DOCUMENT_SUMMARY_CONFIG,
     )
     summary = summary_response.candidates[0].content.parts[0].text
     __indexes__ = [pxt.BtreeIndex(user_id), pxt.BtreeIndex(timestamp)]
@@ -157,27 +179,19 @@ class VideoAudioChunks(
     has_default_idxs=False,
 ):
     transcription = openai.transcriptions(audio=audio, model=config.WHISPER_MODEL_ID)  # type: ignore[name-defined]
-
-
-class VideoTranscriptSentences(
-    TableModel,
-    name="video_transcript_sentences",
-    base=VideoAudioChunks.where(VideoAudioChunks.transcription != None),  # noqa: E711
-    iterator=string_splitter(VideoAudioChunks.transcription.text, separators="sentence"),
-    has_default_idxs=False,
-):
+    text = transcription.text
     __indexes__ = [
         pxt.EmbeddingIndex(text, string_embed=geminiEmbed, name="video_sentences_gemini"),  # type: ignore[name-defined]
     ]
 
 
 @pxt.query
 def search_video_transcripts(query_text: str, user_id: str = config.DEFAULT_USER_ID):
-    sim = VideoTranscriptSentences.text.similarity(string=query_text)
+    sim = VideoAudioChunks.text.similarity(string=query_text)
     return (
-        VideoTranscriptSentences.where((VideoTranscriptSentences.user_id == user_id) & (sim > 0.7))
+        VideoAudioChunks.where((VideoAudioChunks.user_id == user_id) & (sim > 0.7))
         .order_by(sim, asc=False)
-        .select(VideoTranscriptSentences.text, source_video=VideoTranscriptSentences.video, sim=sim)
+        .select(VideoAudioChunks.text, source_video=VideoAudioChunks.video, sim=sim)
         .limit(20)
     )
 
@@ -198,27 +212,19 @@ class AudioChunks(
     has_default_idxs=False,
 ):
     transcription = openai.transcriptions(audio=audio, model=config.WHISPER_MODEL_ID)  # type: ignore[name-defined]
-
-
-class AudioTranscriptSentences(
-    TableModel,
-    name="audio_transcript_sentences",
-    base=AudioChunks.where(AudioChunks.transcription != None),  # noqa: E711
-    iterator=string_splitter(AudioChunks.transcription.text, separators="sentence"),
-    has_default_idxs=False,
-):
+    text = transcription.text
     __indexes__ = [
         pxt.EmbeddingIndex(text, string_embed=geminiEmbed, name="audio_sentences_gemini"),  # type: ignore[name-defined]
     ]
 
 
 @pxt.query
 def search_audio_transcripts(query_text: str, user_id: str = config.DEFAULT_USER_ID):
-    sim = AudioTranscriptSentences.text.similarity(string=query_text)
+    sim = AudioChunks.text.similarity(string=query_text)
     return (
-        AudioTranscriptSentences.where((AudioTranscriptSentences.user_id == user_id) & (sim > 0.6))
+        AudioChunks.where((AudioChunks.user_id == user_id) & (sim > 0.6))
         .order_by(sim, asc=False)
-        .select(AudioTranscriptSentences.text, source_audio=AudioTranscriptSentences.audio, sim=sim)
+        .select(AudioChunks.text, source_audio=AudioChunks.audio, sim=sim)
         .limit(30)
     )
 
@@ -427,6 +433,7 @@ class Notifications(TableModel, name="notifications", has_default_idxs=False):
     search_video_transcripts,
     search_audio_transcripts,
 )
+agentToolDeclarations = _jsonb_stable(agentTools.ser_model())
 
 
 class ToolAgent(TableModel, name="tools", has_default_idxs=False):
@@ -442,7 +449,7 @@ class ToolAgent(TableModel, name="tools", has_default_idxs=False):
     initial_response = gemini.generate_content(
         contents=tool_selection_messages,
         model=config.GEMINI_MODEL_ID,
-        tools=agentTools,
+        tools=agentToolDeclarations,
         config={"system_instruction": initial_system_prompt, "temperature": temperature},
     )
     tool_output = invoke_tools(agentTools, initial_response)
@@ -474,11 +481,7 @@ class ToolAgent(TableModel, name="tools", has_default_idxs=False):
     follow_up_raw_response = gemini.generate_content(
         contents=follow_up_input_message,
         model=config.GEMINI_MODEL_ID,
-        config={
-            "system_instruction": "Generate exactly 3 relevant follow-up questions based on the conversation.",
-            "response_mime_type": "application/json",
-            "response_schema": FollowUpResponse.model_json_schema(),
-        },
+        config=FOLLOW_UP_CONFIG,
     )
     follow_up_text = follow_up_raw_response.candidates[0].content.parts[0].text
     __indexes__ = [pxt.BtreeIndex(user_id), pxt.BtreeIndex(timestamp)]

backend/pixelbot/static/assets/arrow-right-B3HaW9QC.js → backend/pixelbot/static/assets/arrow-right-DbPzFz_k.js

@@ -1 +1 @@
-import{c as o}from"./index-bcIN4R_6.js";const r=[["path",{d:"M5 12h14",key:"1ays0h"}],["path",{d:"m12 5 7 7-7 7",key:"xquz4c"}]],c=o("arrow-right",r);export{c as A};
+import{c as o}from"./index-BKlAZSPs.js";const r=[["path",{d:"M5 12h14",key:"1ays0h"}],["path",{d:"m12 5 7 7-7 7",key:"xquz4c"}]],c=o("arrow-right",r);export{c as A};