crisng95 · Sep 17, 2026 · Sep 15, 2026 · Sep 17, 2026 · Sep 17, 2026 · Sep 17, 2026
diff --git a/.gitignore b/.gitignore
@@ -21,4 +21,6 @@ GEMINI.md
 ds-bundle/
 .design-sync/.cache/
 .design-sync/learnings/
-.design-sync/node_modules
+.design-sync/node_modules
+# Chrome generates this when it loads the unpacked extension
+extension/_metadata/
diff --git a/README.md b/README.md
@@ -16,6 +16,23 @@
 
 ---
 
+> **Working against the new Google Flow API.** Flow moved to `flow.google.com`
+> in September 2026 and stopped minting the `Bearer ya29.…` that the old
+> `aisandbox-pa.googleapis.com` REST API needed. The `batchexecute` transport
+> that replaces it is in place and verified end to end against the live API —
+> image generation, 2K image export, and image-to-video all run green. Upgrading
+> from an older Flow Kit: reload the extension (v0.3.2+) and pin
+> `FLOW_PROJECT_ID`; Flow Kit can no longer create the project for you.
+>
+> Four capabilities are still unported because their payloads were never
+> captured off the new UI — **video upscale**, **reference-to-video (r2v)**,
+> **start+end-frame chaining**, and **Omni Flash frame/reference**. They fail
+> loudly with `UNSUPPORTED_ON_BATCH_API` instead of quietly producing the wrong
+> thing. `FLOW_ALLOW_DEGRADED=1` drops chaining and r2v to plain i2v; video
+> upscale has no fallback. To restore one properly see [`docs/CAPTURE.md`](docs/CAPTURE.md).
+
+---
+
 ### ☕ Sponsor this project
 
 <table align="center">
@@ -188,9 +205,9 @@ One signed-in Flow tab has to stay open; nothing here works headless.
 > **September 2026 — Flow moved.** It now lives at `flow.google.com` and the old
 > `aisandbox-pa.googleapis.com` REST API has no caller: the `Bearer ya29.…` it
 > needed stopped being minted. If you are upgrading from an older Flow Kit,
-> reload the extension (v0.3.0+) and pin `FLOW_PROJECT_ID` — see
-> [Configuration](#configuration). The legacy path is still there behind
-> `USE_BATCH_RPC=0`, but it is a post-mortem tool, not a fallback.
+> reload the extension (v0.3.2+) and pin `FLOW_PROJECT_ID` — see
+> [Configuration](#configuration). The REST path has been removed; `git log`
+> has it if a payload is ever needed for reference.
 
 ## Quick Start
 
@@ -240,7 +257,6 @@ You can also pass `flow_project_id` per project on `POST /api/projects`.
 | Env var | Default | What it does |
 |---------|---------|--------------|
 | `FLOW_PROJECT_ID` | — | The Flow project every RPC is scoped to. Required. |
-| `USE_BATCH_RPC` | `1` | `0` falls back to the pre-migration REST path (dead auth). |
 | `FLOW_ALLOW_DEGRADED` | `0` | `1` lets scene chaining and r2v fall back to plain i2v instead of failing. |
 | `DEFAULT_PAYGATE_TIER` | `PAYGATE_TIER_TWO` | Carried for the DB and dashboard; no longer selects a model. |
 
@@ -690,7 +706,6 @@ agent/
 │                         #   flow, models, providers, reviews, materials, music, tts)
 ├── services/
 │   ├── flow_client.py   # WS bridge to extension
-│   ├── headers.py       # Randomized browser headers
 │   ├── tts.py           # OmniVoice TTS (subprocess-based)
 │   ├── scene_chain.py   # Continuation scene logic
 │   ├── video_reviewer.py # AI vision review — contact sheet + claude/agy/codex CLI dispatch
@@ -805,13 +820,13 @@ These arrive in the response body as `data.error.details[].reason`. The worker a
 | Status | Source | Meaning | Handling |
 |--------|--------|---------|----------|
 | **400** | Flow API | Invalid payload, UNSAFE_GENERATION, entity not found (sometimes) | Route by `details.reason` — some are auto-recoverable, others terminal |
-| **401** | Flow API (legacy path only) | Bearer expired — on a post-migration profile it was never minted | Switch to the batch path (`USE_BATCH_RPC=1`) |
+| **401** | Flow API | Should not occur — batchexecute authenticates in the page, not with a bearer | Check the Flow tab is signed in; see `NO_AT_TOKEN` |
 | **403** | Extension (`background.js:432`) | `CAPTCHA_FAILED`, `NO_FLOW_TAB`, or `MODEL_ACCESS_DENIED` | CAPTCHA → retry loop; NO_FLOW_TAB → fail (user must open Flow); tier → fail |
 | **404** | Flow API | `media_id` not found (expired upload) | Same as "Requested entity was not found" — auto re-upload |
 | **429** | Flow API | Rate limited / quota | Back off + retry; if `USER_QUOTA_REACHED` appears, fail |
 | **500** | Flow backend **or** extension fetch exception (`background.js:504`) | Transient server error OR network drop during fetch | Retry with exponential backoff |
 | **502** | FastAPI default (`agent/api/flow.py:80,92`) | Extension returned error without explicit status | Retry; check extension health |
-| **503** | FastAPI (`api/flow.py`) | "Extension not connected" or `NO_FLOW_KEY` | Worker waits for reconnect — status set to PENDING, not FAILED |
+| **503** | FastAPI (`api/flow.py`) | "Extension not connected" | Worker waits for reconnect — status set to PENDING, not FAILED |
 | **504** | Agent | 60s timeout waiting for extension WS response | Treated as transient; re-queue PENDING |
 
 Status-code detection logic lives in `agent/worker/_parsing.py:_is_error` — a result is an error if `result.error` is set, `status >= 400`, **or** `data.error` is present.
@@ -825,7 +840,6 @@ String patterns in `error_message` that the worker recognizes:
 | `Extension not connected` | Chrome extension offline or WS dropped | 503 returned; worker re-queues PENDING and waits |
 | `extension reconnected` / `extension disconnected` | WS bounce mid-request | Re-queue PENDING without incrementing `retry_count` |
 | `extension_switched` | User switched Flow tabs mid-generation | Re-queue PENDING |
-| `NO_FLOW_KEY` | Extension has no captured bearer token — **legacy path only**, expected on the batch path | Only meaningful with `USE_BATCH_RPC=0` |
 | `NO_AT_TOKEN` | Flow tab is signed out, on an interstitial, or still booting | Open `flow.google.com`, sign in, let the app load |
 | `NO_FLOW_PROJECT` | No Flow project to scope the RPC to | Pin `FLOW_PROJECT_ID` — **terminal, not retried** |
 | `UNSUPPORTED_ON_BATCH_API` | Upscale / r2v / chaining — payload never captured | See `docs/CAPTURE.md` — **terminal, not retried** |

diff --git a/agent/api/flow.py b/agent/api/flow.py
@@ -3,7 +3,7 @@
 from pydantic import BaseModel, Field
 from typing import Literal, Optional
 
-from agent.config import USE_BATCH_RPC, FLOW_PROJECT_ID, FLOW_ALLOW_DEGRADED
+from agent.config import FLOW_PROJECT_ID, FLOW_ALLOW_DEGRADED
 from agent.services.flow_client import get_flow_client
 from agent.services.omni_flash import (
     check_omni_flash_status,
@@ -122,15 +122,17 @@ class UpscaleImageRequest(BaseModel):
 
 @router.get("/status")
 async def extension_status():
-    """Extension health, and which transport it is being asked to speak.
+    """Extension health.
 
-    `flow_key_present` is a legacy-path signal: the batchexecute path has no
-    bearer token at all, so false is expected there rather than a fault.
+    `flow_key_present: false` is expected, not a fault: batchexecute
+    authenticates in the page and there is no bearer token to capture.
     """
     client = get_flow_client()
     return {
         "connected": client.connected,
-        "transport": "batch" if USE_BATCH_RPC else "legacy_rest",
+        # One transport now. The key stays so the documented pre-flight check
+        # (CLAUDE.md) keeps reading {"transport": "batch", ...}.
+        "transport": "batch",
         "flow_project_id": FLOW_PROJECT_ID or None,
         "allow_degraded": FLOW_ALLOW_DEGRADED,
         "flow_key_present": client._flow_key is not None,

diff --git a/agent/api/projects.py b/agent/api/projects.py
@@ -7,7 +7,7 @@
 from fastapi import APIRouter, HTTPException
 from pydantic import BaseModel
 
-from agent.config import BASE_DIR, USE_BATCH_RPC
+from agent.config import BASE_DIR
 from agent.models.project import Project, ProjectCreate, ProjectUpdate
 from agent.models.character import Character
 from agent.sdk.persistence.sqlite_repository import SQLiteRepository
@@ -172,9 +172,9 @@ async def create(body: ProjectCreate):
 
     detected_tier = await _detect_user_tier(client)
 
-    # On the batch path Flow no longer creates projects for us — the uuid comes
-    # from the request or from FLOW_PROJECT_ID. The legacy path still mints one.
-    flow_project_id = client.flow_project_id(body.flow_project_id) if USE_BATCH_RPC else None
+    # Flow no longer creates projects for us — the uuid comes from the request
+    # or from FLOW_PROJECT_ID.
+    flow_project_id = client.flow_project_id(body.flow_project_id)
     if flow_project_id:
         logger.info("Flow project reused: %s", flow_project_id)
     else:

diff --git a/agent/config.py b/agent/config.py
@@ -15,19 +15,12 @@
 WS_HOST = os.environ.get("WS_HOST", "127.0.0.1")
 WS_PORT = int(os.environ.get("WS_PORT", "9222"))
 
-# ─── Google Flow API ────────────────────────────────────────
-# Legacy REST host. Flow moved to flow.google.com in September 2026 and stopped
-# minting the `Bearer ya29.…` this host needs, so these are only reachable with
-# USE_BATCH_RPC=0 on a browser profile that still has an old token.
-GOOGLE_FLOW_API = "https://aisandbox-pa.googleapis.com"
-GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY", "AIzaSyBtrm0o5ab1c-Ec8ZuLcGt3oJAA5VWt3pY")
-RECAPTCHA_SITE_KEY = os.environ.get("RECAPTCHA_SITE_KEY", "6LdsFiUsAAAAAIjVDZcuLhaHiDn5nnHVXVRQGeMV")
-
-# ─── Flow batchexecute (the current path) ───────────────────
+
+# ─── Flow batchexecute ──────────────────────────────────────
 # Every call is signed in the page with the session cookie plus a per-page `at`
-# token, so the extension runs it inside a signed-in flow.google.com tab. Set
-# USE_BATCH_RPC=0 only to fall back to the dead REST path for a post-mortem.
-USE_BATCH_RPC = os.environ.get("USE_BATCH_RPC", "1") == "1"
+# token, so the extension runs it inside a signed-in flow.google.com tab. This
+# is the only transport; the REST path it replaced was removed once Flow stopped
+# minting the bearer it needed.
 
 # The Flow project every RPC is scoped to. Project creation went with the old
 # labs.google tRPC endpoint, so a project is made once in the Flow UI and its
@@ -66,20 +59,6 @@
 # newly introduced image models do not require a Flow Kit release.
 DEFAULT_IMAGE_MODEL = _MODELS.get("default_image_model", "NANO_BANANA_PRO")
 
-# ─── API Endpoints ───────────────────────────────────────────
-ENDPOINTS = {
-    "generate_images": "/v1/projects/{project_id}/flowMedia:batchGenerateImages",
-    "generate_video": "/v1/video:batchAsyncGenerateVideoStartImage",
-    "generate_video_start_end": "/v1/video:batchAsyncGenerateVideoStartAndEndImage",
-    "generate_video_references": "/v1/video:batchAsyncGenerateVideoReferenceImages",
-    "upscale_video": "/v1/video:batchAsyncGenerateVideoUpsampleVideo",
-    "upscale_image": "/v1/flow/upsampleImage",
-    "upload_image": "/v1/flow/uploadImage",
-    "check_video_status": "/v1/video:batchCheckAsyncVideoGenerationStatus",
-    "get_credits": "/v1/credits",
-    "get_media": "/v1/media/{media_id}",
-}
-
 # ─── Output Directories ─────────────────────────────────────
 OUTPUT_DIR = BASE_DIR / "output"
 SHARED_OUTPUT_DIR = OUTPUT_DIR / "_shared"
@@ -131,29 +110,3 @@ def _load_suno_key() -> str:
 SUNO_POLL_INTERVAL = int(os.environ.get("SUNO_POLL_INTERVAL", "5"))
 SUNO_POLL_TIMEOUT = int(os.environ.get("SUNO_POLL_TIMEOUT", "600"))
 
-# ─── Header Randomization Pools ─────────────────────────────
-USER_AGENTS = [
-    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/109.0.0.0 Safari/537.36",
-    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36",
-    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/111.0.0.0 Safari/537.36",
-    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/109.0.0.0 Safari/537.36",
-    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36",
-    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/111.0.0.0 Safari/537.36",
-]
-
-CHROME_VERSIONS = [
-    '"Google Chrome";v="109", "Chromium";v="109"',
-    '"Google Chrome";v="110", "Chromium";v="110"',
-    '"Google Chrome";v="111", "Chromium";v="111"',
-    '"Google Chrome";v="113", "Not-A.Brand";v="24"',
-    '"Google Chrome";v="120", "Not-A.Brand";v="24"',
-    '"Google Chrome";v="141", "Not?A_Brand";v="8", "Chromium";v="141"',
-]
-
-BROWSER_VALIDATIONS = [
-    "SgDQo8mvrGRdD61Pwo8wyWVgYgs=",
-]
-
-CLIENT_DATA = [
-    "CKi1yQEIh7bJAQiktskBCKmdygEIvorLAQiUocsBCIagzQEYv6nKARjRp88BGKqwzwE=",
-]
diff --git a/agent/sdk/services/operations.py b/agent/sdk/services/operations.py
@@ -40,7 +40,7 @@ def _char_matches(c: dict, name_set: set) -> bool:
 import aiohttp
 
 from agent.db import crud
-from agent.config import USE_BATCH_RPC, VIDEO_POLL_INTERVAL, VIDEO_POLL_TIMEOUT
+from agent.config import VIDEO_POLL_INTERVAL, VIDEO_POLL_TIMEOUT
 from agent.utils.paths import scene_4k_path
 from agent.utils.slugify import slugify
 from agent.worker._parsing import (
@@ -447,28 +447,19 @@ async def generate_scene_video(self, scene: dict, orientation: str,
             base_prompt = scene.get("video_prompt") or scene.get("prompt", "")
         prompt = await _build_video_prompt(base_prompt, scene, pid)
 
-        # Check if already submitted (op_name saved from previous attempt)
-        # OLD schema (Lite/Fast/Ultra): op_name is "models/.../operations/..." → re-poll via check_video_status
-        # NEW schema (Low Priority workflow): op_name is bare UUID → cannot recover (need primary_media_id
-        # which isn't persisted yet); fall through and resubmit (Low Priority is free, duplicate is OK)
+        # Already submitted on a previous attempt? Re-poll it.
         existing_op = None
         if request_id:
             req_row = await crud.get_request(request_id)
             existing_op = req_row.get("request_id") if req_row else None
 
-        # A bare uuid means different things on the two transports. On the
-        # legacy path it is a Low Priority workflow name that cannot be
-        # re-polled, so the retry resubmits. On the batch path it is the
-        # operation id, and looking it up in the project listing is exactly
-        # what the status poll does — resubmitting there would abandon a
-        # running render and pay for a second one.
-        bare_uuid = bool(existing_op and len(existing_op) == 36 and existing_op.count("-") == 4)
-        looks_like_workflow_uuid = bare_uuid and not USE_BATCH_RPC
-        if existing_op and not looks_like_workflow_uuid:
+        # A bare uuid is the operation id, and looking it up in the project
+        # listing is exactly what the status poll does — resubmitting instead
+        # would abandon a running render and pay for a second one.
+        if existing_op:
             logger.info("Video gen already submitted (op=%s), re-polling", existing_op[:30])
             operations = [{"operation": {"name": existing_op}, "status": "MEDIA_GENERATION_STATUS_PENDING"}]
             return await _poll_operations(self._client, operations)
-        # else: workflow UUID — fall through and resubmit fresh
 
         submit_result = await self._client.generate_video(
             start_image_media_id=image_media_id,