feat: add Sats4AI recipes — visual asset creator + Goosetown flock#8291
feat: add Sats4AI recipes — visual asset creator + Goosetown flock#8291cnghockey wants to merge 3 commits intoaaif-goose:mainfrom
Conversation
cnghockey
requested review from
DOsinga,
angiejones and
blackgirlbytes
as code owners
|
🔍 Recipe Security Scan Results 📊 Scan Summary:
📋 Individual Recipe Results: 🔗 View detailed scan results in the workflow artifacts. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 23b21172de
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| "url": "https://sats4ai.com/api/mcp", | ||
| "link": "https://sats4ai.com/mcp", | ||
| "is_builtin": false, | ||
| "endorsed": false |
There was a problem hiding this comment.
This new server object omits environmentVariables, but the install-link path assumes it is always an array and immediately calls .filter(...) in documentation/src/utils/install-links.ts when rendering install buttons. For this non-built-in server, opening the extensions list/detail will hit getGooseInstallLink(server) and throw at runtime because server.environmentVariables is undefined.
Useful? React with 👍 / 👎.
| Do not retry with the same paymentId — create a new payment for the retry. | ||
|
|
||
| ## MCP server setup (if not already configured) | ||
| Add the Sats4AI MCP server to Goose: |
There was a problem hiding this comment.
The documentation style guide in documentation/AGENTS.md requires the product name to always be written as lowercase goose across documentation content. This new recipe text uses Goose, which violates that repository guideline and should be corrected for consistency with the enforced brand standard.
Useful? React with 👍 / 👎.
cnghockey
changed the title
|
🔍 Recipe Security Scan Results 📊 Scan Summary:
📋 Individual Recipe Results: 🔗 View detailed scan results in the workflow artifacts. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c2cd8968e7
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
|
|
||
| Follow these steps: | ||
|
|
||
| 1. **Check pricing first** — call list_models and find the current prices for generate_image and remove_background. Report them before proceeding. |
There was a problem hiding this comment.
This recipe calls Sats4AI tools (list_models, create_payment, generate_image, etc.) but never defines a corresponding extensions entry, so the generated goose://recipe config is built with no Sats4AI extension attached; in a fresh goose setup, launching this recipe will fail on the first tool invocation unless the user manually pre-configured the server. Adding a streamable_http extension stanza for https://sats4ai.com/api/mcp would make the launch path functional by default.
Useful? React with 👍 / 👎.
| @@ -0,0 +1,193 @@ | |||
| version: 1.0.0 | |||
| title: "Research & Visual Content Flock (Sats4AI + Goosetown)" | |||
| description: Orchestrate a parallel flock of Goose agents — researchers gather context, workers generate images and remove backgrounds simultaneously — all paid with Bitcoin Lightning via Sats4AI. | |||
There was a problem hiding this comment.
documentation/AGENTS.md requires the product name to always be written as lowercase goose throughout this directory, but this new recipe text introduces Goose in user-facing content, which violates the documented style rule and creates inconsistent branding in published docs.
Useful? React with 👍 / 👎.
2 participants
What this adds
Two recipes using Sats4AI — a Bitcoin-powered AI tools marketplace (remote streamable-http MCP, no install, no API keys).
Recipe 1: AI Visual Asset Creator
Single-agent workflow. User provides a subject → agent generates N image variations in parallel → removes backgrounds from all of them in parallel.
Demonstrates the batch-pay-then-parallel-execute pattern: all Lightning payments created upfront, user pays once, all tools execute simultaneously.
Parameters: subject, style (5 options), count (1/2/4), remove_bg (yes/no)
Recipe 2: Research & Visual Content Flock (Goosetown)
Multi-agent orchestration using Goosetown (https://github.com/block/goosetown). The only recipe in the gallery that demonstrates Goosetown's flock coordination pattern with external paid tools.
Phase 1 — Research flock: spawns 1–3 parallel researcher delegates (visual references, context/mood, trends). Agents coordinate via gtwall.
Phase 2 — Synthesize: orchestrator reads gtwall, crafts N distinct image prompts.
Phase 3 — Batch payments: creates ALL Lightning micropayments upfront (image gen + background removal for every worker), shows all invoices at once.
Phase 4 — Worker flock: spawns N parallel worker delegates, each claiming their gtwall slot, generating an image, then removing the background.
Phase 5 — Aggregate: orchestrator collects final URLs from gtwall, delivers complete asset set + total sats spent.
Parameters: topic, image_count (1/2/4), style (5 options), research_depth (1/2/3 researchers)
MCP server: streamable-http at
https://sats4ai.com/api/mcp— no install, no authDocs: https://sats4ai.com/mcp