update ideogram4 requirements and global skills update

Signed-off-by: Vladimir Mandic <mandic00@live.com>

.github/instructions/core.instructions.md

@@ -1,8 +1,22 @@
 ---
 description: "Use when editing Python core runtime code, startup flow, model loading, API internals, backend/device logic, or shared state in modules and pipelines."
 name: "Core Runtime Guidelines"
-applyTo: "launch.py, webui.py, installer.py, modules/**/*.py, pipelines/**/*.py, scripts/**/*.py, extensions-builtin/**/*.py"
+applyTo: "launch.py, webui.py, installer.py, modules/**/*.py, pipelines/**/*.py, scripts/**/*.py, extensions-builtin/**/*.py, cli/**/*.py"
 ---
+## Agent Guidelines
+
+1. Verify the user instruction against relevant guidelines in this file and linked instruction files before proceeding.
+2. If the instruction conflicts with any guideline, do not proceed. Explain which guideline(s) it conflicts with and how to adjust the instruction to comply.
+3. If the instruction is valid but unclear or incomplete, ask targeted follow-up questions before implementation. Do not assume user intent or requirements.
+4. When giving feedback, name the applicable guideline(s) and explain how each one applies.
+5. If the instruction is clear and compliant, proceed and keep resulting changes aligned with project coding style, conventions, and structure.
+
+## Language Guidelines
+
+- Use clear and concise language when communicating with users, providing feedback, and explaining guidelines.
+- Avoid unnecessary pleasantries or filler language; focus on the technical content and actionable feedback.
+- When asking follow-up questions for clarification, be direct and specific about the information needed to proceed with the instruction while ensuring that the questions are relevant to the project guidelines and conventions.
+
 # Core Runtime Guidelines
 
 1. Preserve startup ordering and import timing in `launch.py` and `webui.py`; avoid moving initialization steps unless required to fix a critical startup bug or implement a new startup feature.
@@ -14,6 +28,16 @@ applyTo: "launch.py, webui.py, installer.py, modules/**/*.py, pipelines/**/*.py,
 7. Reuse established model-loading and pipeline patterns (`modules/sd_*`, `pipelines/`) instead of creating parallel abstractions.
 8. For substantial Python changes, run at least relevant checks: `pnpm run ruff` and `pnpm run pylint` (or narrower equivalents when appropriate).
 
+## Tools
+
+- `venv` for Python environment management, activated with `source venv/bin/activate` (Linux) or `venv\Scripts\activate` (Windows).  
+  venv MUST be activated before running any Python commands or scripts to ensure correct dependencies and environment variables.  
+- `python` 3.10+.
+- `pyproject.toml` for Python configuration, including linting and type checking settings.
+- `pnpm` for managing JavaScript dependencies and scripts, with key commands defined in `package.json`.
+- `ruff` and `pylint` for Python linting, with configurations in `pyproject.toml` and executed via `pnpm ruff` and `pnpm pylint`.
+- `pre-commit` hooks which also check line-endings and other formatting issues, configured in `.pre-commit-config.yaml`.
+
 ## Build And Test
 
 - Activate environment: `source venv/bin/activate` (always ensure this is active when working with Python code).
@@ -21,7 +45,6 @@ applyTo: "launch.py, webui.py, installer.py, modules/**/*.py, pipelines/**/*.py,
 - Full startup: `python launch.py`
 - Full lint sequence: `pnpm lint`
 - Python checks individually: `pnpm ruff`, `pnpm pylint`
-- TypeScript checks: `pnpm eslint`, `pnpm tsc`
 
 ## Pitfalls
 

.github/instructions/ui.instructions.md

@@ -3,6 +3,20 @@ description: "Use when editing frontend UI code, TypeScript, JavaScript, HTML, C
 name: "UI And Frontend Guidelines"
 applyTo: "ui/**/*, extensions-builtin/sdnext-modernui/**/*, extensions-builtin/sdnext-kanvas/**/*"
 ---
+## Agent Guidelines
+
+1. Verify the user instruction against relevant guidelines in this file and linked instruction files before proceeding.
+2. If the instruction conflicts with any guideline, do not proceed. Explain which guideline(s) it conflicts with and how to adjust the instruction to comply.
+3. If the instruction is valid but unclear or incomplete, ask targeted follow-up questions before implementation. Do not assume user intent or requirements.
+4. When giving feedback, name the applicable guideline(s) and explain how each one applies.
+5. If the instruction is clear and compliant, proceed and keep resulting changes aligned with project coding style, conventions, and structure.
+
+## Language Guidelines
+
+- Use clear and concise language when communicating with users, providing feedback, and explaining guidelines.
+- Avoid unnecessary pleasantries or filler language; focus on the technical content and actionable feedback.
+- When asking follow-up questions for clarification, be direct and specific about the information needed to proceed with the instruction while ensuring that the questions are relevant to the project guidelines and conventions.
+
 # UI And Frontend Guidelines
 
 Apply these rules in priority order:

.github/skills/analyze-model/SKILL.md

@@ -14,6 +14,10 @@ Given an external model URL, inspect how the model is implemented and estimate h
 - User wants effort estimation before implementation work
 - User wants to classify whether integration should reuse Diffusers, use custom Diffusers code, or require full custom implementation
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and model integration guidance before proceeding.
+
 ## Accepted Inputs
 
 - Hugging Face model URL (preferred)

.github/skills/check-api/SKILL.md

@@ -15,6 +15,10 @@ Read modules/api/api.py, enumerate all registered endpoints, and validate that e
 - OpenAPI docs look wrong or clients report schema mismatches
 - You need a pre-PR API contract sanity pass
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and API guidance before proceeding.
+
 ## Primary File
 
 - `modules/api/api.py`

.github/skills/check-models/SKILL.md

@@ -15,6 +15,10 @@ Run a consolidated model-integration audit that combines loader checks, detect/r
 - A custom pipeline was ported and needs contract validation
 - You want a pre-PR integration quality gate for model-related changes
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and model integration guidance before proceeding.
+
 ## Combined Scope
 
 This skill combines four audit surfaces. Run them in this order unless user scope limits categories:

.github/skills/check-processing/SKILL.md

@@ -15,6 +15,10 @@ Perform a detailed step-by-step trace of generation workflows from UI definition
 - A new parameter was added to UI or processing classes/functions and needs end-to-end validation
 - You want a pre-PR contract audit for generation flow integrity
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## Required Workflow Coverage
 
 Run workflow coverage in this order to keep checks focused and complete:

.github/skills/check-schedulers/SKILL.md

@@ -29,6 +29,10 @@ Related files:
 - `modules/schedulers/**/*.py` for custom scheduler implementations
 - `modules/res4lyf/**/*.py` for Res4Lyf scheduler classes (if installed/enabled)
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## What "Loaded And Compiled" Means
 
 Treat this as a two-level check:

.github/skills/check-scripts/SKILL.md

@@ -15,6 +15,10 @@ Audit all Python scripts in `scripts/*.py` and validate that script class overri
 - A script UI was changed and runtime args no longer match
 - You want a pre-PR quality gate for script API compatibility
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## Scope
 
 Primary audit scope:

.github/skills/check-ui/SKILL.md

@@ -15,6 +15,10 @@ Audit SD.Next UI integration points where Python uses Gradio `_js=...` bindings
 - You need to validate UI contract consistency for Gradio-bound JS methods.
 - User adds or updates extension with JavaScript code in `extensions/*/javascript`.
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## Primary Files
 
 - `ui/globals.d.ts`

.github/skills/debug-model/SKILL.md

@@ -17,6 +17,10 @@ Read the error, identify which integration layer is failing, isolate the smalles
 - Sampling fails due to tensor shape, dtype, device, or scheduler issues
 - The model loads but outputs corrupted images, wrong output type, or obviously incorrect results
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and model debugging guidance before proceeding.
+
 ## Debugging Order
 
 Always debug from the outside in.

.github/skills/diffusers-code/SKILL.md

@@ -16,6 +16,10 @@ Use this skill to implement, edit, review, and prepare pull-request-ready change
 - Adding tests and docs for diffusers changes
 - Preparing a PR that targets the diffusers repository
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and diffusers integration guidance before proceeding.
+
 ## Primary Objectives
 
 1. Keep behavior explicit, minimal, and inference-focused.

.github/skills/fix-lint/SKILL.md

@@ -15,6 +15,10 @@ Run the project lint workflow in the required order, fix findings, and re-run af
 - Multiple files changed and style/static checks may have drifted
 - You need a repeatable full-lint remediation pass
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## Required Environment Step
 
 Always start from repository root and activate virtual environment first:

.github/skills/port-model/SKILL.md

@@ -16,6 +16,10 @@ Read the task, identify the model architecture and artifact layout, choose the i
 - A model already exists in Diffusers but is not yet wired into SD.Next
 - A custom architecture needs a repo-local `pipelines/<model>` package and loader
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and model porting guidance before proceeding.
+
 ## Core Rule
 
 Prefer the smallest correct integration path.

.github/skills/port-pipeline/SKILL.md

@@ -17,6 +17,10 @@ This skill targets SD.Next repo-local pipeline ports only.
 - The task requires preserving generation behavior without introducing new dependencies
 - The task requires removing hard-coded runtime assumptions (device or attention backend)
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime and pipeline integration guidance before proceeding.
+
 ## Mandatory Clarification Gate
 
 Before implementation, confirm these required inputs with the user:

.github/skills/reference-catalog/SKILL.md

@@ -15,6 +15,10 @@ Use this skill to audit and update SD.Next model reference catalogs using a phas
 - Verifying category placement across `base/cloud/quant/distilled/nunchaku/community`
 - Syncing catalog entries with thumbnail files in `models/Reference`
 
+## Guidance
+
+- Consult `.github/instructions/core.instructions.md` for relevant core runtime guidance before proceeding.
+
 ## Catalog Files In Scope
 
 - `data/reference.json` (base)
@@ -40,7 +44,7 @@ Priority 2 - schema and formatting consistency:
 Priority 3 - assets and size backfill:
 
 - Do not overwrite real thumbnails with placeholders.
-- For `size` backfill, use `cli/hf-info.py` as the primary source of truth.
+- For `size` backfill, use `cli/hf-info.py` -> section `info` -> field `size` as the primary source of truth.
 
 ## Validation Checklist
 
@@ -76,7 +80,7 @@ Priority 3 - assets and size backfill:
 7. Size backfill checks (`size: 0`)
 - Enumerate all entries with `"size": 0` across `data/reference*.json`.
 - For each Hugging Face repo-style path (`owner/name`), run `cli/hf-info.py`.
-- Parse `data.size` from tool output when present (format is MB string, e.g. `"23933.4MB"`).
+- Parse `info.data.size` from tool output when present (format is MB string, e.g. `"23933.4MB"`).
 - Convert MB to GB using deterministic rounding: `gb = round(mb / 1024, 2)`.
 - Update only the `size` field for resolvable records; do not modify unrelated fields.
 - If `cli/hf-info.py` returns `ok: false`, missing `data.size`, or non-repo paths, leave `size` unchanged and report as unresolved.

installer.py

@@ -533,7 +533,7 @@ def check_diffusers():
     t_start = time.time()
     if args.skip_all:
         return
-    target_commit = "9b0818cf87413b4b9ca2501bf49406eed6d881af" # diffusers commit hash == 0.39.0.dev0 == 06-03-2026 (adds Ideogram 4)
+    target_commit = "79e408a705d6aff9762c58e4601b11e7ea00ea86" # diffusers commit hash == 0.39.0.dev0 == 06-03-2026 (adds Ideogram 4)
     # if args.use_rocm or args.use_zluda or args.use_directml:
     #     sha = '043ab2520f6a19fce78e6e060a68dbc947edb9f9' # lock diffusers versions for now
     pkg = package_spec('diffusers')

pipelines/generic_shared.py

@@ -1,6 +1,5 @@
 import os
 import transformers
-from transformers.models.qwen3_vl import Qwen3VLModel
 
 
 shared_te_map = {
@@ -98,7 +97,7 @@ shared_te_map = {
     },
 
     'Qwen3-VL 8B Base': {
-        'cls': Qwen3VLModel,
+        'cls': transformers.Qwen3VLModel,
         'target_repo': 'Qwen/Qwen3-VL-8B-Instruct',
     },
 }

pipelines/model_ideogram4.py

@@ -42,7 +42,7 @@ class Ideogram4Pipeline(diffusers.Ideogram4Pipeline):
                 self.text_encoder.to(devices.cpu)
 
 
-def pin_transformers_if_fit(transformer, unconditional_transformer) -> bool:
+def pin_transformers(transformer, unconditional_transformer) -> bool:
     """Keep both transformers resident under balanced offload when they fit the budget.
 
     Every denoise step runs both transformers, so balanced offload ping-pongs them across
@@ -60,7 +60,7 @@ def pin_transformers_if_fit(transformer, unconditional_transformer) -> bool:
     if fits:
         transformer.offload_never = True
         unconditional_transformer.offload_never = True
-    log.info(f'Load model: type=Ideogram4 offload=balanced transformers={size_gb:.1f} budget={budget_gb:.1f} action={"pin-resident" if fits else "offload"}')
+    log.debug(f'Load model: type=Ideogram4 offload=balanced transformers={size_gb:.1f} budget={budget_gb:.1f} action={"pin" if fits else "default"}')
     return fits
 
 
@@ -80,7 +80,7 @@ def load_ideogram4(checkpoint_info, diffusers_load_config=None):
     cls = diffusers.Ideogram4Transformer2DModel
     transformer = generic.load_transformer(repo_id, cls_name=cls, subfolder="transformer", load_config=diffusers_load_config)
     unconditional_transformer = generic.load_transformer(repo_id, cls_name=cls, subfolder="unconditional_transformer", load_config=diffusers_load_config)
-    pin_transformers_if_fit(transformer, unconditional_transformer)
+    pin_transformers(transformer, unconditional_transformer)
     # shared_te_map redirects to the shared Qwen3-VL repo (deduped with VQA + prompt-enhance);
     # the bundled text_encoder is the fallback when sharing is off. The vae, tokenizer, and
     # scheduler load from the repo via from_pretrained.