OUTCOME DELIVERY PLATFORM
Describe the outcome. Flushnet builds, automates and delivers it.
Use a simple task for one direct action, or Build & Deliver when you want Flushnet to plan and execute a larger outcome.
Workspace status
Ready
Sign in to see your active work
START SOMETHING
What would you like to do?
Choose the outcome; advanced execution details stay out of the way until you need them.
LIVE BUILDER SHOWCASE
See what Flushnet can build before you start
Explore production-style websites, business apps, dashboards, presentations, games and AI agents created with Flushnet App Builder, Presentation Builder and Agent Builder. Every public demo includes the exact prompt or specification, so you can copy it, reproduce the result and adapt it for your own project.
RUNNING WORK
In progress
Sign in to view running work.
NEEDS ATTENTION
Blockers and failures
Sign in to view items needing attention.
RECENTLY COMPLETED
Latest finished work
Sign in to view completed work.
DELIVERABLES
Ready to use
Sign in to view deliverables.
SIMPLE TASK
Run one direct task
Choose what you want done, provide the input, and Flushnet handles the execution details.
Checking which services can run automatically now…
Estimated cost10 Compute credits
MANAGED DELIVERY
Complex job request
Describe the complete outcome you need. Flushnet will assess feasibility, estimate cost and delivery time, and accept the job only when it can be scheduled.
ACTIVITY
Job history
| ID | Type | Status | Cost | Created | Progress | Result / Actions |
|---|
PROJECTS
Persistent workspaces
A project keeps the objective, execution plan, jobs, progress and final results together over time.
AI APP / MOBILE BUILDER
Build a web or mobile app from scratch
Describe the product in plain language. Flushnet generates source files, validates builds, previews supported web targets and deploys or exports the result. AI agents are built separately in Agent Builder.
READY TO BUILD
Your AI development workspace
Generate an app to open its files, edit code, request changes, preview it and deploy it.
Select a file
Versions & deployments
Managed database
Environment secrets
Custom domain
Build backends
GitHub is optional. Apps can still be generated and use the build, validation, preview, export and deployment capabilities available on Flushnet web resources.
GitHub Actions builds optional
No GitHub account is required to build apps. Connect GitHub only when you want its external Actions runners and artifact pipeline.
Use access token instead
Use a fine-grained token with repository contents, workflows and Actions access.
Pipeline: sync source → send GitHub Action → compile/validate → monitor → save to web workspace or download to your computer.
Build and deployment logs
No logs yet.
AI PRESENTATION BUILDER
Build PowerPoint presentations from a prompt
Create the outline, slide copy and browser preview with Flushnet AI, then render real PowerPoint and OpenDocument files through the Flushnet MS Office / LibreOffice backend.
PRESENTATION WORKSPACE
Browser preview + real Office exports
Generate a deck to preview it here. PPTX is rendered by the Flushnet Office backend; ODP is converted through LibreOffice. Backend PDF export is currently unavailable and is not advertised as working.
Export backend · PowerPoint uses the Flushnet MS Office/python-pptx renderer. ODP uses LibreOffice 7.6. PDF is hidden until the current LibreOffice PDF-export issue is repaired.
AUDIO TRANSCRIPTION
Turn audio and video into editable transcripts
Transcribe locally in the browser, with Flushnet Hosted STT, or through managed high-accuracy STT. Browser/device STT keeps the source media on your device and uses 0 AI Credits; Auto falls back to Flushnet Hosted STT when needed and never enables a paid route silently.
Drop an audio or video file hereor click to browse · browser STT can process large local files without uploading them
Or record audio directly in your browser.
Browser STT · Checking browser capability…
Smart recovery · Used only after Hosted/managed STT when the primary transcript is empty, unusually short, repetitive, low-confidence, or fails. Browser-only transcription never silently starts a paid model.
Workflow estimateSelect an audio fileBrowser/device STT: 10 Compute credits + 1 per started minute, with 0 AI Credits and no source upload. Server routes retain the existing file-size floor. Managed paid STT requires explicit confirmation.
Large files · Browser/device STT reads the selected local File through the browser media decoder and processes it in sequential 4-minute recognition windows; the source file is not uploaded to Flushnet, so the 120 MB server-upload ceiling does not apply. Processing is approximately real-time and depends on browser/media support. If Auto cannot use browser STT, Hosted fallback is available only for files within the normal direct-upload limit; larger failures can use GPU Compute or be split locally.
Your transcripts
Browser-local, hosted and managed transcription results.
RAG / KNOWLEDGE BASE
Turn documents into an AI-ready knowledge base
Extraction, chunking, semantic embeddings, indexing and retrieval are managed as one production workflow. Use the result from Flushnet AI, Agent Builder, App Builder or your own API.
Embedding routing
Auto tries verified-free semantic embedding routes first and can fall back to Flushnet-managed paid embeddings when needed. Free only never falls back to a paid model. BYOK uses your configured provider account and consumes 0 Flushnet AI Credits. Managed uses normal Flushnet AI Credit accounting.
RAG processing/indexing is a Compute service charge and is separate from embedding-provider inference. A free embedding route therefore does not make document processing free.
Use this knowledge everywhere
Agent Builder
Bind this account knowledge base to an agent.Open
Bind this account knowledge base to an agent.Open
App Builder
Use the RAG API from a generated app.
Use the RAG API from a generated app.
Stable binding
flushnet://rag/default
flushnet://rag/default
Plan-aware limits
Sign in to load your current Free / Pro / Power limits.
Free: small manual knowledge base. Pro: larger indexed corpus and user-connected storage sync. Power: high-volume corpus, larger files and higher sync limits. Original files stay in the user's connected storage; Flushnet does not resell shared-hosting disk as cloud storage. Provider costs still follow the same free / managed / BYOK accounting rules on every plan.
ETL / ELT SERVICES
Open advanced ETL studioBuild reliable data pipelines
Extract, clean, transform, enrich and load data into warehouses, databases, APIs or files.
ETL / ELT pipeline
Checking automation path…
Your ETL/ELT requests
GPU COMPUTE
External accelerated compute, routed by Flushnet
Request GPU / AI compute for inference, media, Python/CUDA, training and batch workloads. Flushnet selects an eligible execution backend internally, so the customer workflow stays provider-neutral.
Run a workload for me
Describe the task and required compute. Flushnet reviews the workload and routes eligible jobs to managed accelerated compute.
- AI inference and batch processing
- Audio / transcription workloads
- Embeddings and data workloads
- No upstream provider account required in the customer workflow
Request external GPU capacity
For heavier Python/CUDA, model-training or longer-running workloads, request a compute class and duration. Capacity and price are confirmed before paid provisioning.
- Provider-neutral compute classes
- Single or multi-accelerator requests
- Preferred region
- No paid capacity starts without confirmation
New GPU Compute request
Managed GPU compute · request/quote workflow
Request submission0 Compute creditsExternal GPU usage is billed separately after capacity, price and execution scope are confirmed. No paid capacity is started by this button.
Your GPU Compute requests
Track quote/activation status here. Normal execution jobs also appear in Job history.
MODEL TRAINING
Open advanced training studioTrain custom AI and machine-learning models
Use your own Hugging Face account, or let Flushnet sponsor provider costs and charge the actual cost plus commission.
Hugging Face account
Tokens are encrypted and never displayed after saving.
Flushnet-sponsored Hugging Face jobs
Customers can request permission to use the admin Hugging Face account. Approved jobs are charged at the actual provider cost plus the approved Flushnet commission.
Pending sponsorship approvals
Model training
Checking production training availability…
Your training requests
APPS & AUTOMATIONS
What Flushnet has built for you
FILES
Workspace files
Uploaded inputs and generated workspace artifacts. Final user-facing outputs are kept separately in Deliverables.
| Name | Type | Size | Added | Action |
|---|
INFRASTRUCTURE
Runtime resources
Workers, services, databases and deployed resources used by your projects.
DELIVERABLES
Finished results
Final outputs ready to use, download or open, linked back to the project that produced them when available.
ADVANCED OPERATIONS
AI processing queue
Execution diagnostics for advanced users and administrators. Normal task progress is shown in Jobs and on the Dashboard.
BILLING
Credits & model costs
Compute execution and AI provider inference are accounted separately.
Workflow and build execution
Used for Flushnet Compute tasks, app-building, presentation-building, audio-transcription workflow processing, ETL, training services and managed execution where shown.
Managed paid model inference
Used only when a managed paid AI model/provider route is selected and admitted by Flushnet Router. This is separate from Compute Credits.
0 AI Credits
Verified free model routes consume no AI Credits. Compute workflow charges can still apply to the surrounding task or build.
0 Flushnet AI Credits
Your provider account pays the provider directly. Flushnet does not charge AI Credits for that provider inference.
Buy Compute Credits
These packs add to your Flushnet Compute execution balance.
€10
1,000 Compute Credits
€49
5,500 Compute Credits
€99
12,000 Compute Credits
DEVELOPER
API keys
Use API keys for the standard Flushnet APIs enabled for your account.
Super-admin external App / Agent / Presentation Builder control
Internal integration for verified super admins only. Use this with Flushnet Commander, AI Runner custom ChatGPT, or another trusted admin AI client to control App Builder, Agent Builder and Presentation Builder.
One-command build: https://www.flushnet.net/flushnet-compute/api/build.php
Unified control: https://www.flushnet.net/flushnet-compute/api/external_control.php
OpenAPI: https://www.flushnet.net/flushnet-compute/api/builder-actions-openapi.json
| Name | Prefix | Created | Last used |
|---|
OPERATIONS
Admin control plane
Worker nodes
| Name | Status | Capabilities | Last seen |
|---|
Users
| Role | Credits | Status | Actions |
|---|
Backend execution router
Internal only. Select which backend process and optional model handles each operation. Customers only see status and charges. Generate/iterate execute through the selected configured provider adapter when available; the existing Compute credit admission and provider-usage accounting remain in force, with controlled NVIDIA fallback if the selected provider is unavailable.
External AI provider connections
Credentials are encrypted and never displayed after saving. These connections are used by App Builder generation/iteration when selected by a routing rule. For AWS, put the Secret Access Key in Credential and the Access Key ID below; temporary/session credentials can instead be stored as a JSON object in the Credential field. Azure requires the resource endpoint and deployment/model name. Google Cloud uses the Gemini API-key endpoint. External API must be a public HTTPS OpenAI-compatible endpoint.
Not tested
NVIDIA internal AI
Admin-managed. Used by backend routing when enabled.
Checking…