• project - cofounder.openinterface.ai
• 👋 @n_raidenai

cofounder

• full stack generative web apps ; backend + db + stateful web apps
• gen ui rooted in app architecture, with ai-guided mockup designer & modular design systems

The following points are very emphasized :

• This is an EARLY, UNSTABLE, PREVIEW RELEASE of the project. ⚠️ Until v1 is released, it is expected to break often.
• It consumes a lot of tokens. If you are on a tokens budget, wait until v1 is released.
• Again, this is an early, unstable release. A first test run. An early preview of the project's ideas. Far from completion. Open-source iterative development. Work in progress. Unstable early alpha release. [etc]

To help you guide your decision on whether or not to try the current release , here is a guide

Early alpha release ; earlier than expected by few weeks

Still not merged with key target features of the project, notably :

• project iteration modules for all dimensions of generated projects
• admin interface for event streams and (deeper) project iterations
• integrate the full genUI plugin :
  • generative design systems
  • deploy finetuned models & serve from api.cofounder
• local, browser-based dev env for the entire project scope
• add { react-native , flutter , other web frameworks }
• validations & swarm code review and autofix
• code optimization
• [...]

be patient :)

• Open your terminal and run

npx @openinterface/cofounder

Follow the instructions. The installer

• will ask you for your keys
• setup dirs & start installs
• will start the local cofounder/api builder and server
• will open the web dashboard where you can create new projects (at http://localhost:4200 ) 🎉

note :
you will be asked for a cofounder.openinterface.ai key
it is recommended to use one as it enables the designer/layoutv1 and swarm/external-apis features
and can be used without limits during the current early alpha period

the full index will be available for local download on v1 release

• currently using node v22 for the whole project.

# alternatively, you can make a new project without going through the dashboard
# by runing :
npx @openinterface/cofounder -p "YourAppProjectName" -d "describe your app here" -a "(optional) design instructions"

• Your backend & vite+react web app will incrementally generate inside ./apps/{YourApp} Open your terminal in ./apps/{YourApp} and run

npm i && npm run dev

It will start both the backend and vite+react, concurrently, after installing their dependencies Go to http://localhost:5173/ to open the web app 🎉

• From within the generated apps , you can use ⌘+K / Ctrl+K to iterate on UI components

[more details later]

If you resume later and would like to iterate on your generated apps, the local ./cofounder/api server needs to be running to receive queries

You can (re)start the local cofounder API running the following command from ./cofounder/api

npm run start

The dashboard will open in http://localhost:4200

• note: You can also generate new apps from the same env, without the the dashboard, by running, from ./cofounder/api, one of these commands

  npm run start -- -p "ProjectName" -f "some app description" -a "minimalist and spacious , light theme"
  npm run start -- -p "ProjectName" -f "./example_description.txt" -a "minimalist and spacious , light theme"

[the architecture will be further detailed and documented later]

Every "node" in the cofounder architecture has a defined configuration under ./cofounder/api/system/structure/nodes/{category}/{name}.yaml to handle things like concurrency, retries and limits per time interval

For example, if you want multiple LLM generations to run in parallel (when possible - sequences and parallels are defined in DAGS under ./cofounder/api/system/structure/sequences/{definition}.yaml ), go to

#./cofounder/api/system/structure/nodes/op/llm.yaml
nodes:
 op:LLM::GEN:
  desc: "..."
  in: [model, messages, preparser, parser, query, stream]
  out: [generated, usage]
  queue:
   concurrency: 1 # <------------------------------- here 
 op:LLM::VECTORIZE:
  desc: "{texts} -> {vectors}"
  in: [texts]
  out: [vectors, usage]
 mapreduce: true
 op:LLM::VECTORIZE:CHUNK:
  desc: "{texts} -> {vectors}"
  in: [texts]
  out: [vectors, usage]
  queue:
   concurrency: 50

and change the op:LLM::GEN parameter concurrency to a higher value

The default LLM concurrency is set to 2 so you can see what's happening in your console streams step by step - but you can increment it depending on your api keys limits

• | Community discord server by @flamecoders

[WIP]

[more details later]

archi/v1 is as follows :

• Demo design systems built using Figma renders / UI kits from:
  • blocks.pm by Hexa Plugin (see cofounder/api/system/presets)
  • google material
  • figma core
  • shadcn
• Dashboard node-based ui powered by react flow