## Summary

Two community finetunes of Anima now install and generate correctly:
[Anima-2.9B](https://huggingface.co/Gazingstars123/Anima-2.9B) and
[Anima-3.8B](https://huggingface.co/lylogummy/Anima-3.8B).

**Anima-2.9B** (40 DiT blocks instead of 28) already installed as Anima,
but the loader built the transformer from a hard-coded 28-block config
and loaded with `strict=False`. It filled the first 28 blocks and
dropped the other 12 (240 tensors) as unexpected keys, logged at DEBUG
only. The result was a degraded image and no error. The loader now reads
the depth from the checkpoint and refuses gaps. The int8_tensorwise
build from the same repo loads too, through
`install_int8_convrot_layers`.

**Anima-3.8B v1.1** (52 blocks) bundles a "semantic connector" into its
checkpoint that reads a second text encoder, Qwen3.5 4B. The connector
is timestep-aware, so its context changes at every step. Until now the
checkpoint installed as Anima, loaded 28 blocks and dropped the
connector. Now:

- `AnimaVariantType` (`anima_qwen3`, `anima_qwen35`), detected from the
bundled `anima_v2_connector.*` keys
- new `ModelType.Qwen35Encoder` (`qwen3_5_encoder`): single-file config
and loader built on the `transformers` `qwen3_5` decoder layers, with
the tokenizer vendored from `Qwen/Qwen3.5-4B` @ `851bf6e` (Apache-2.0).
The encoder checkpoint is not a stock export: it ships layer 31 without
its MLP, and the encoder reproduces that exactly.
- a port of the connector
(`invokeai/backend/anima/semantic_connector.py`), whose module names
match the checkpoint
- **Prompt - Anima** encodes Qwen3.5 when its new optional input is
connected. **Denoise - Anima** recomputes the connector context at every
step, for the positive, the negative and every regional conditioning,
with a float32 sigma. The connector scales sigma by 1000 before
embedding it, and bf16 rounding would shift its high frequencies.
- webv2: a **Qwen3.5 Encoder** slot shown only for the `anima_qwen35`
variant, plus graph, regional guidance and recall wiring
- the earlier Anima-3.8B v1.0 transformer, which needs a separate
adapter file, is refused at install with an `InvalidMatchError` that
names v1.1

**LoRAs and ControlNet-LLLite adapters.** Both finetunes insert their
new blocks between the original ones, so an adapter trained on Anima
patched different blocks from the third one on. Measured from the
checkpoints:
- Anima-2.9B holds all 28 blocks of Anima base-v1.0 bit for bit, with
its 12 new blocks at 2, 5, 8, …, 36.
- Anima-3.8B holds the 40 blocks of 2.9B nearly unchanged (cosine
similarity ≥ 0.9997, 11 bit-identical), with 12 new blocks at 3, 7, …,
47.

`invokeai/backend/anima/block_layout.py` records the tables. A LoRA's
DiT keys and an LLLite adapter's bindings now move to the blocks of the
model they were trained on: Anima adapters on 2.9B and 3.8B, and 2.9B
adapters on 3.8B. The adapter's depth is read off the highest block it
addresses.

Also included:
- starter models for both finetunes (2.9B bf16 and int8, 3.8B v1.1) and
the Qwen3.5 encoder, with the non-commercial license noted
- the Anima user guide page
- three additions to `new-model-integration.mdx`, from the gaps this
change hit:
  - a new section, *Extending an existing architecture*
  - a section on persisted records and migrations
  - corrections to the webv2 checklists

<!-- Screenshots: the Components section with the Qwen3.5 Encoder slot
for Anima-3.8B, and sample outputs. -->

## Related Issues / Discussions

- Builds on #9613 (merged).
- #9506 (Anima prompt weighting) still targets the v6 path
`invokeai/app/invocations/anima_text_encoder.py`. In v7 the node lives
at `invokeai/app/invocations/text_encoder/anima_text_encoder.py`, and
this PR changes it as well: the new Qwen3.5 input, and the conditioning
fields it stores. #9506 needs a v7 port either way, and that port picks
up this file.
- Reference implementation:
[GumGum10/comfyui-anima-3-8B](https://github.com/GumGum10/comfyui-anima-3-8B)
(MIT).

## QA Instructions

**Automated**
- `uv run pytest -n 8 tests/backend/model_manager
tests/backend/architectures tests/backend/anima tests/backend/qwen3_5
tests/backend/util tests/backend/patches tests/app/invocations
tests/app/services/shared tests/app/services/model_records
tests/test_package_data.py`: 6254 passed, 1 failed. The failure,
`test_16_channel_vae_loader.py::test_an_ldm_layout_file_is_converted_and_loses_no_tensor`,
reads an LFS fixture that was not pulled in the test worktree; it is
unrelated.
- `ruff check .` and `ruff format --check .`: clean.
- webv2: `lint:tsc`, `lint:oxc`, `format:check` and `architecture:check`
pass. Vitest over `src/features/generation`, `src/workbench`,
`src/features/models` and `src/features/workflow`: 7277 passed, 2
failed. Both failures are in image-map (`clusterStats`, `indexProgress`)
and come from German-locale digit grouping on the test machine;
unrelated.
- Regenerated: `openapi.json`/`schema.ts`, the capabilities fixture, and
the graph-coverage snapshot. The docs build passes, and
`check-docs-data` shows no diff.

**After review**
- **Startup import:** the prompt node imports the Qwen3.5 encoder where
it runs, not at module level. That spares every app start transformers'
`qwen3_5` modeling module, about 0.25 s. A subprocess test fails on the
previous module-level import.
- **Migration on real databases:** the full chain ran on copies of three
populated databases, using `init_db` as the app does, and every model
record read back afterwards.
- A v7 dev root with Anima base, preview and a tcfp8 build: all three
became `anima_qwen3`.
  - A v7 dev root with Anima base fp8: it became `anima_qwen3`.
- The production database of a v6 install, 69 MiB with 75 models: the
chain ran in 0.3 s and all 75 records read back. It holds no Anima main
model.
- Two Ideogram-4 GGUF records from another branch's build were invalid
before and after; they are unrelated.
- **FP8 Storage and the connector's timestep path** (Anima-3.8B,
832×1216, 30 steps, three seeds; every run fully resident and
bit-reproducible):

  | | PSNR against bf16 |
  |---|---|
| timestep path kept in bf16 (the previous exception) | 8.4 / 10.3 / 9.6
dB |
  | everything in FP8 | 8.4 / 10.2 / 9.6 dB |
  | whole connector in bf16 | 8.3 / 10.1 / 9.6 dB |
  | Anima base, FP8 against bf16 | 15.9 / 15.2 / 14.0 dB |

The exception changed nothing measurable and cost 152 MiB (159M
parameters), so it is gone. FP8 Storage changes Anima-3.8B's composition
through its DiT blocks, not the connector; the user guide now says so.
- **LLLite through the block layout** (`anima-lllite-any-test-like-v2`,
trained on Anima). Control: a grayscale render. Prompt: different
content. The metric is the edge correlation between output and control:

| | no adapter | adapter by index (before) | adapter on the kept blocks
(now) |
  |---|---|---|---|
  | Anima base | 0.126 | 0.451 | 0.451 |
  | Anima-2.9B | 0.121 | 0.187 | 0.474 |
  | Anima-3.8B | 0.081 | 0.106 | 0.421 |

By index, the finetunes all but ignored the adapter; on the kept blocks
they follow it as Anima base does. The LoRA path shares the tables and
is unit-tested (key moves, untouched LLM adapter and text encoder keys,
cached LoRA not modified); no Anima LoRA was available locally for an
end-to-end run.

<!-- Screenshots: LLLite before/after sheet, FP8 Storage sheet for
Anima-3.8B. -->

**Numerical checks against the reference implementation, on the real
weights**
- Qwen3.5 encoder, fp32: relative L2 ~1.5e-6 at layers 7/15/23/31, for
27- and 114-token prompts. In bf16 the error is ~1e-2, the same as the
reference's own bf16 error.
- Connector: bit-identical (relative L2 0.0) at sigma 1.0, 0.75, 0.3 and
0.02. An empty prompt (one masked Qwen3.5 token) also matches the
reference, on both the math and the efficient SDPA backends.

**End to end**: RTX 4090, fresh root, models installed in place through
the API, 832×1216, 30 steps, Euler, seed 1234.

| Model | Result |
| --- | --- |
| Anima Base 1.0 | unchanged (regression check) |
| Anima-2.9B bf16 | coherent, all 40 blocks loaded |
| Anima-2.9B int8 | nearly identical to bf16; 640 of 640 layers kept
int8, 2.9 GB in VRAM instead of 5.6 GB |
| Anima-3.8B v1.1, CFG 6 | coherent, with clearly better spatial and
object binding |
| Anima-3.8B, empty negative prompt | coherent |
| Anima-3.8B without a Qwen3.5 encoder | refused by the model loader
with a message naming the missing encoder |

Speed, single runs only (not a benchmark): ~1.7 it/s for 2.9B against
~1.3 it/s for 3.8B, which also runs the connector every step.

**Browser** (built webv2 against the same server):
- The Qwen3.5 slot appears only when Anima-3.8B is selected. Invoke
stays disabled until both encoders are chosen.
- Generation works, and the image metadata records `qwen3_5_encoder`.
- **Reset all to model defaults** sets 40 steps / CFG 6, the settings of
the author's reference workflow.


## Review

Remaining risks and limitations:
- The block layout tables are keyed by depth (40, 52). These are the
only depth-expanded Anima models today. A future finetune with one of
these depths but another layout would need its own entry.
- An adapter's own depth is read off the highest block it addresses. An
adapter trained on Anima-2.9B that addresses only blocks below 28 would
be taken for an Anima adapter.
- On Anima-3.8B, the blocks an Anima adapter lands on were changed
slightly, and the semantic connector conditions every block. Adapters
work there (see the LLLite table), but results can differ more than on
2.9B.
- The reference masks every Qwen3.5 token from the first id 151643 on.
That id is the Qwen3 padding token, but in Qwen3.5's vocabulary it is
the ordinary token " 내용". This behavior is not reproduced; empty prompts
behave like the reference.
- The author's workflow samples with `res_multistep` and a beta
schedule, which Anima's scheduler set does not offer.
- The int8 path is new for Anima and has been validated on the
Anima-2.9B int8 file only.

## Compatibility / Rollout

- **Migration** `2026_10_01_add_anima_variant` (depends on
`2026_09_26_add_workflow_revision`): `variant` is now a required field
on Anima main records. Without the migration, every installed Anima
model would fail validation on read and vanish from the model list. The
migration reads each checkpoint's header, so a 3.8B installed on an
earlier v7 build becomes `anima_qwen35`. A record whose file cannot be
read gets `anima_qwen3`.
- **New persisted values**: model type `qwen3_5_encoder`, variants
`anima_qwen3`, `anima_qwen35` and `qwen3_5_4b`.
- **Node versions**: `anima_model_loader` 1.5.0 and `anima_text_encoder`
1.5.0. The new inputs and outputs are optional, so existing workflows
keep validating.
- **Packaging**: `invokeai.backend.qwen3_5` is added to `package-data`.
A new `tests/test_package_data.py` checks that every vendored
`*.json`/`*.json.gz` under `invokeai/backend` ships in the wheel.

## Checklist

- [x] _The PR has a short but descriptive title, suitable for a
changelog_
- [x] _Meaningful regression coverage added / updated where needed;
obsolete tests/code removed_
- [x] _Persisted-state and API changes include required migrations /
compatibility validation_
- [x] _Relevant performance/efficiency opportunities considered;
material claims have evidence_
- [x] _Material review findings resolved and relevant checks rerun_
- [x] _Documentation added / updated (if applicable)_
- [ ] _Updated `What's New` copy (if doing a release after this PR)_

…te (#9644)

## Summary

This brings back the "What's New" notes from 6.14.2 and earlier in the
v7 (webv2) UI, redesigned around the v7 highlights, and makes the UI
show the real release version instead of a hard-coded `7.0`.

- **What's New dialog.** A larger dialog led by the v7 logo, the title
and a server-version badge. Headline features show as icon cards
(`whatsNew.highlights`: redesigned interface, projects, rebuilt Canvas,
expanded model support), with the remaining notes listed under "Also in
this release" (`whatsNew.items`). **Read the Docs** and **Read Release
Notes** (`releases/tag/v<server version>`) are footer buttons. The
`<StrongComponent>` emphasis from 6.14 is gone; card titles carry that
emphasis now.
- **Automatic first-run showing.** The dialog opens on its own once per
server version for each account. It waits until the account's workbench
preferences have loaded and the alpha notice has been dismissed, so the
two never stack. Dismissing it stores `whatsNewSeenVersion` in the
account's preferences.
- **Lightbulb button.** A lightbulb sits in the app menu footer, between
Settings and Documentation. It uses the same Phosphor
`LightbulbFilament` (bold) glyph as 6.14.2, vendored into
`VendoredIcon.tsx` because webv2 doesn't depend on react-icons.
- **Launchpad entry.** The launchpad has no app menu, so its sidebar
footer gets a "What's New in Invoke" action with the same lightbulb,
between Preferences and Help and resources.
- **Home in the Invoke menu.** The editor's Invoke menu gains a **Home**
item, with the launchpad's house icon, above the Manage group.
- **App version.** `APP_VERSION` was hard-coded to `'7.0'` and went
stale (the server is `7.0.0-rc1`). Vite now reads `__version__` from
`invokeai/version/invokeai_version.py`, the same file the release
workflow checks against the tag. The splash screen, app menu, Help menu,
`.invk` export manifest, and diagnostics and log exports all show the
real version. The release process still bumps only that one file.
- **Docs link.** The app-menu Documentation icon and the launchpad Help
→ Documentation entry now point to
https://invoke-ai.github.io/InvokeAI-7/.

### Implementation notes
- **Where the server version lives.** The server version is kept in
`whatsNewStore`, not in the TanStack query cache. The gate mounts above
the router, and account transitions call `queryClient.clear()` without
remounting it. A query-backed version therefore never arrived in the
real app, even though the unit tests passed. The version is loaded from
the authenticated route's `beforeLoad`, and opening the dialog from the
menu retries it.
- **Closing before the version loads.** A manual close sets a dismissed
flag for the rest of the page load. Without it, a version that arrived
late would reopen notes the user had just closed.
- **Initial focus.** Focus starts on the dialog panel. An automatic open
has no click behind it, so focusing the first link would have drawn a
keyboard focus ring on it.
- **Highlights catalog.** Each card's icon is paired with its
`whatsNew.highlights.<key>` entry in `WhatsNewDialog.tsx`, so adding a
headline feature means adding both. The browser test fails if a catalog
highlight has no card.
- **Build-time version.** The version is injected as
`import.meta.env.APP_VERSION`, not as a global `__APP_VERSION__` define:
Vitest browser mode re-encodes string globals, so they arrive wrapped in
an extra pair of quotes. The build fails if the version file is missing
or its `__version__` line can't be parsed; there is no fallback.
- **Startup bundle.** `appMetadata`, `whatsNewStore` and `useWhatsNew`
join `ROUTE_SHARED_MODULES`, so startup request counts stay at the
baseline. The build and browser baselines are re-recorded for those
three modules entering the startup graph (measured cost under QA
Instructions). The baseline's image-byte drop comes from main's assets
changing since the 2026-09-29 capture. The 113 KB logo loads only when
the dialog opens.
- **Translation-key test.** `translationKeys.test.ts` now treats an
array's own path as a key, because arrays are read whole with
`returnObjects`.
- **Test fixtures.** The mock backend seeds `whatsNewSeenVersion` so
journeys don't open on the dialog, and the topbar-menu accessibility
journey includes the new footer item in its keyboard order.

## Related Issues / Discussions

Follows #9613 (InvokeAI 7 alpha); the highlights are drawn from its
release notes.

## QA Instructions

- `pnpm lint` (format, oxlint, tsc, architecture): passes.
- `pnpm test`: 9290 passed.
- `pnpm test:browser`: all files pass, including `WhatsNewDialog` tests
for:
- automatic open, persisting the version on dismissal, and alpha-notice
gating;
  - every catalog highlight and note rendering;
  - manual open, which writes nothing;
  - a late version after a manual close;
  - a new account seeing the notes again;
  - initial focus.
- `pnpm test:accessibility`: all 17 journeys pass.
- `pnpm test:performance:build` (`check-architecture-performance.mjs`):
passes.
- `pnpm test:performance:browser`: passes after re-recording the browser
baseline for the three What's New boot modules. Startup scripts grow by
4.8 KB on the launchpad and 7.5 KB in the editor, and other assets by
1.2 KB (mostly the English catalog), with no new requests. The editor's
image bytes drop by 112 KB because the previous capture predates main's
asset changes. The 113 KB logo is not loaded at startup.
- App version: `APP_VERSION` is `7.0.0-rc1` in Node Vitest, browser
Vitest, `vite dev` and the production bundle. A build in the Docker
stage's directory layout succeeds, and fails with a clear error when the
version file is missing or malformed.
- Not run: the Docker image build itself.
- In a browser against the mock backend (dark and light themes; 1440×900
and 1280×720):
- the dialog opens automatically for an unseen version and has no focus
ring;
  - after dismissing it and reloading, it stays closed;
  - the menu button opens it;
  - Tab reaches the footer links and Esc closes it;
  - at 1280×720 the body scrolls under a fixed header and footer.

## Review

No material findings remain. Remaining limitations:
- **Tab with an outdated version.** A tab left open since before an
upgrade pushes its whole preference document whenever any setting
changes. That can roll `whatsNewSeenVersion` back, so the notes show
once more. Every preference already behaves this way.
- **Late version.** If the version request is slow, the notes can appear
on top of another dialog that is already open.
- **Settings-load fallback.** When the backend settings load fails and
local settings are used instead, the dialog can close and reopen on
navigation.
- **Dev builds.** The release-notes link points to a tag that doesn't
exist on development builds.
- **Anima models.** The "Expanded model support" card mentions larger
Anima models, which have not merged yet.

## Compatibility / Rollout

- **Docker.** The web-builder stage now mirrors the repository layout
(`/build/invokeai/frontend/webv2` plus
`/build/invokeai/version/invokeai_version.py`), and the final stage
copies `dist` from the new path.
- **Builds outside the repository.** A webv2 build without
`invokeai/version/invokeai_version.py` two directories up now fails. The
wheel script, CI workflows and Dockerfile all build inside the
repository layout.

## Checklist

- [x] _The PR has a short but descriptive title, suitable for a
changelog_
- [x] _Meaningful regression coverage added / updated where needed;
obsolete tests/code removed_
- [ ] _Persisted-state and API changes include required migrations /
compatibility validation_
- [x] _Relevant performance/efficiency opportunities considered;
material claims have evidence_
- [x] _Material review findings resolved and relevant checks rerun_
- [x] _Documentation added / updated (if applicable)_
- [x] _Updated `What's New` copy (if doing a release after this PR)_

🤖 Generated with [Claude Code](https://claude.com/claude-code)