Default provider → Anthropic Claude Opus 4.7; GitHub Pages landing

Provider
- Default LLM is now Anthropic Claude Opus 4.7 (was MiniMax M2)
- AGUI_LLM_BASE_URL defaults to https://api.anthropic.com
- AGUI_LLM_MODEL defaults to claude-opus-4-7
- ANTHROPIC_API_KEY is now the primary fallback env var
- All docs, bin/huxform[.ps1], README, env.example updated
- MiniMax still works — just flip the four AGUI_LLM_* vars

GitHub Pages
- docs/index.html — a stage-first landing that demonstrates HUXForm
  by behaving like one: hero with rotating verb, before/after diptych,
  specimen cabinet with real PNGs of generated UIs, architecture,
  one-command quickstart, footer manifesto
- docs/assets/ — banner, sigil, architecture, real screenshots
- docs/.nojekyll — serve HTML as-is
- Self-contained: one HTML file, Google Fonts, no build step
- Enable by setting Pages source to "main / docs" in repo settings

Screenshots (assets/)
- Re-generated with English-only prompts and richer scenarios:
  · what's the weather in New York → meteorological_station_readout
    (deep navy METAR readout, 72°F NYC, condition cards)
  · explain MCP → mcp_technical_manual
    (cream paper, draft v1.0 — 2024 badge, justified spec text)
  · compare three payment processors for a US-based SaaS
    → payment_processor_scorecard (Stripe/Braintree/Adyen with real
       fees, scores, per-row Verdict badges)
  · find duplicates in this CSV → duplicate_forensics_bench

Repo hygiene
- idea.md untracked, added to .gitignore
- Dev playwright screenshots in repo root are ignored

Author: pavvvel13722-alt

.env.example

@@ -1,11 +1,10 @@
 # --- LLM provider --------------------------------------------------------
 # HUXForm talks to any provider that speaks the Anthropic Messages API
-# or the OpenAI Chat Completions API. Default: MiniMax M2 via its
-# Anthropic-compatible endpoint.
+# or the OpenAI Chat Completions API. Default: Anthropic Claude Opus 4.7.
 AGUI_LLM_PROTOCOL=anthropic
-AGUI_LLM_BASE_URL=https://api.minimax.io/anthropic
-AGUI_LLM_MODEL=MiniMax-M2
-AGUI_LLM_API_KEY=your_minimax_key_here
+AGUI_LLM_BASE_URL=https://api.anthropic.com
+AGUI_LLM_MODEL=claude-opus-4-7
+AGUI_LLM_API_KEY=your_anthropic_key_here
 AGUI_LLM_MAX_TOKENS=4096
 AGUI_LLM_TEMPERATURE=0.6
 

.gitignore

@@ -15,6 +15,20 @@ dist/
 .omc/
 test-dupes.csv
 apps/web/test-dupes.csv
+idea.md
+
+# Dev screenshots from playwright runs (keep curated ones in assets/)
+/shot-*.png
+/page-*.png
+/landing-*.png
+/payment-*.png
+/csv-*.png
+/stage-*.png
+/sessions-*.png
+/huxform-*.png
+/page-hero.png
+/page-manifesto.png
+/page-cabinet.png
 .huxform-api.log
 .huxform-web.log
 *.log

README.md

@@ -47,22 +47,22 @@ moment.
   <table>
     <tr>
       <td align="center" width="50%">
-        <img src="./assets/shot-explainer.png" alt="Museum specimen card explainer" width="100%"/>
-        <sub><b>explainer · technical specimen card</b><br/>"explain what MCP is" — cream parchment, serif display, taxonomy ledger</sub>
+        <img src="./assets/shot-weather.png" alt="Meteorological station readout" width="100%"/>
+        <sub><b>status_view · meteorological station readout</b><br/>"what's the weather in New York?" — cold observatory palette, METAR string</sub>
       </td>
       <td align="center" width="50%">
-        <img src="./assets/shot-stage.png" alt="Payment system specimen cabinet" width="100%"/>
-        <sub><b>decision_board · payment specimen cabinet</b><br/>"compare three payment processors" — side-by-side cards, real data</sub>
+        <img src="./assets/shot-mcp.png" alt="MCP technical manual" width="100%"/>
+        <sub><b>explainer · technical manual</b><br/>"explain Model Context Protocol" — paper specimen, justified spec text</sub>
       </td>
     </tr>
     <tr>
       <td align="center" width="50%">
-        <img src="./assets/shot-csv.png" alt="Specimen sorting bench for CSV dedup" width="100%"/>
-        <sub><b>generated_app · specimen sorting bench</b><br/>"check this CSV for duplicates" — working drop-zone bound to the tool broker</sub>
+        <img src="./assets/shot-payments.png" alt="Payment processor scorecard" width="100%"/>
+        <sub><b>decision_board · processor scorecard</b><br/>"compare payment processors" — feature matrix, scores, per-row verdicts</sub>
       </td>
       <td align="center" width="50%">
-        <img src="./assets/shot-landing.png" alt="HUXForm landing" width="100%"/>
-        <sub><b>landing · the moment of stillness</b><br/>one prompt, rotating hint, no chat history, no dashboard chrome</sub>
+        <img src="./assets/shot-csv.png" alt="Duplicate forensics bench" width="100%"/>
+        <sub><b>generated_app · duplicate forensics bench</b><br/>"find duplicates in this CSV" — real drop-zone bound to the tool broker</sub>
       </td>
     </tr>
   </table>
@@ -75,7 +75,7 @@ Same runtime, four completely different surfaces — picked and built per task.
 ## Quick start
 
 Clone the repo and run one script. It checks your Python and Node versions,
-creates a virtualenv, installs dependencies, prompts for your MiniMax key
+creates a virtualenv, installs dependencies, prompts for your Anthropic key
 (or any Anthropic-compatible / OpenAI-compatible key), starts both servers
 and opens your browser.
 
@@ -103,7 +103,7 @@ That's it. The script does the rest:
 
   setup
   ✓  python3 / node / npm preflight
-  paste your MiniMax API key  ⟶  ······
+  paste your Anthropic API key  ⟶  ······
   ✓  wrote .env
   api · creating Python venv
   api · installing dependencies
@@ -127,11 +127,11 @@ skipped automatically.
 | Python       | 3.11+   | `python3 --version` |
 | Node.js      | 20+     | `node --version` |
 | npm          | 10+     | bundled with Node |
-| An LLM key   | —       | MiniMax M2 by default; or any Anthropic / OpenAI-compatible endpoint |
+| An LLM key   | —       | Anthropic Claude Opus 4.7 by default; or any Anthropic / OpenAI-compatible endpoint |
 
-You can grab a free MiniMax key in five minutes at
-[platform.minimax.io](https://platform.minimax.io/). To switch providers,
-edit `.env` after first run (see [Provider configuration](#provider-configuration)).
+Get a key from [console.anthropic.com](https://console.anthropic.com/). To
+switch to MiniMax, OpenAI, OpenRouter, Groq, Together, Ollama or any other
+provider, edit `.env` after first run (see [Provider configuration](#provider-configuration)).
 
 ### Other ways to run
 
@@ -225,8 +225,8 @@ HUXForm is provider-agnostic. Pick the protocol your provider speaks:
 | Var                   | Default                              | Notes                                  |
 |-----------------------|--------------------------------------|----------------------------------------|
 | `AGUI_LLM_PROTOCOL`   | `anthropic`                          | `anthropic` (Messages) or `openai`     |
-| `AGUI_LLM_BASE_URL`   | `https://api.minimax.io/anthropic`   | Provider base URL.                     |
-| `AGUI_LLM_MODEL`      | `MiniMax-M2`                         | Model id.                              |
+| `AGUI_LLM_BASE_URL`   | `https://api.anthropic.com`          | Provider base URL.                     |
+| `AGUI_LLM_MODEL`      | `claude-opus-4-7`                    | Model id.                              |
 | `AGUI_LLM_API_KEY`    | —                                    | API key.                               |
 | `AGUI_LLM_MAX_TOKENS` | `4096`                               |                                        |
 | `AGUI_LLM_TEMPERATURE`| `0.6`                                |                                        |
@@ -236,9 +236,9 @@ HUXForm is provider-agnostic. Pick the protocol your provider speaks:
 | `AGUI_ENABLE_CLI`     | unset                                | Enables `cli.*` host-CLI tools.        |
 | `AGUI_CLI_ALLOWLIST`  | unset                                | Optional `:`-separated allowlist.      |
 
-> Default is **MiniMax M2** via its Anthropic-compatible endpoint. Switch to
-> Anthropic / OpenAI / Groq / OpenRouter / Together / Ollama by changing the
-> four `AGUI_LLM_*` vars — no SDK changes required.
+> Default is **Anthropic Claude Opus 4.7**. Switch to MiniMax, OpenAI, Groq,
+> OpenRouter, Together or Ollama by changing the four `AGUI_LLM_*` vars —
+> no SDK changes required.
 
 ---
 

apps/api/src/config.py

@@ -1,10 +1,9 @@
 """Runtime configuration loaded from environment variables.
 
-AGUI is provider-agnostic. By default we talk to MiniMax M2.7 via its
-Anthropic-compatible endpoint, but any provider speaking either the
-Anthropic Messages or OpenAI Chat Completions protocol will work — just
-point AGUI_LLM_BASE_URL at it and pick the protocol via
-AGUI_LLM_PROTOCOL.
+HUXForm is provider-agnostic. By default we talk to Anthropic Claude
+Opus 4.7, but any provider speaking either the Anthropic Messages or
+OpenAI Chat Completions protocol will work — just point
+AGUI_LLM_BASE_URL at it and pick the protocol via AGUI_LLM_PROTOCOL.
 """
 
 from __future__ import annotations
@@ -34,9 +33,9 @@ def load_llm_config() -> LLMConfig:
     protocol = (_env("AGUI_LLM_PROTOCOL", "anthropic") or "anthropic").lower()
 
     if protocol == "anthropic":
-        base_url = _env("AGUI_LLM_BASE_URL", "https://api.minimax.io/anthropic") or ""
-        api_key = _env("AGUI_LLM_API_KEY") or _env("ANTHROPIC_AUTH_TOKEN") or _env("ANTHROPIC_API_KEY") or _env("MINIMAX_API_KEY") or ""
-        model = _env("AGUI_LLM_MODEL", "MiniMax-M2") or "MiniMax-M2"
+        base_url = _env("AGUI_LLM_BASE_URL", "https://api.anthropic.com") or ""
+        api_key = _env("AGUI_LLM_API_KEY") or _env("ANTHROPIC_API_KEY") or _env("ANTHROPIC_AUTH_TOKEN") or ""
+        model = _env("AGUI_LLM_MODEL", "claude-opus-4-7") or "claude-opus-4-7"
     elif protocol == "openai":
         base_url = _env("AGUI_LLM_BASE_URL", "https://api.openai.com/v1") or ""
         api_key = _env("AGUI_LLM_API_KEY") or _env("OPENAI_API_KEY") or ""

apps/api/src/llm.py

@@ -4,7 +4,7 @@
 send a system prompt + a list of role/content messages, get back text.
 We support the two protocols that nearly every provider speaks:
 
-  * Anthropic Messages API  (POST /v1/messages)        — MiniMax, Anthropic
+  * Anthropic Messages API  (POST /v1/messages)        — Anthropic, MiniMax
   * OpenAI Chat Completions (POST /v1/chat/completions) — OpenAI, OpenRouter,
                                                           Groq, Together, ...