A Deterministic Workflow Compiler, Not a Black Box
Text block, file, or natural-language prompt in — form UI, REST API, and document template out
Self-host on your own infrastructure. No mandatory cloud, no GPU, no phone-home calls
Natural Language to Workflow Optional Layer
This is a convenience layer only. Your prompt is expanded into a tagged text block by an LLM (hosted or a local model like Ollama/Gemma), then that text block is run through the exact same deterministic compiler as Text-to-Workflow below — no separate code path, no hallucinated output logic.
- .
- .
File-to-Workflow
Drop in a DOCX, DOT, RTF, or PDF that's already tagged (or let Auto Tag detect fields for you). The compiler reads the file's text block and deterministically emits a matching form UI, REST API, and document template.
Text-to-Workflow Core Compiler Path
The reference path: a plain text block with {{tagged_fields}} — or fields auto-detected from free text like an email address or phone number — compiles into a form UI, REST API, and document template. Every other path (file upload, natural language) reduces to this same text block before compilation.
How the Compiler Works
Every generation path above collapses to the same three-stage pipeline. Only Stage 1 varies; Stages 2 and 3 are identical, deterministic, and repeatable every time.
1. Input Varies
A tagged text block, an uploaded/tagged file (DOCX, DOT, RTF, PDF), or a natural-language prompt. If a prompt is used, an LLM (hosted or local) expands it into a tagged text block — this is the only step where a model is in the loop.
2. Text Block Always
A normalized, tagged representation such as {{First_Name_req}} or <<field_name>>. This is the single canonical intermediate form the compiler understands — see the tag reference below.
3. Compiled Output Deterministic
Form UI, REST API, and document template are emitted directly from the text block by rule-based logic — no model involved, no variance between runs, no hallucinated fields or validation rules.
Tag Syntax Quick Reference
Wrap every field in {{ }}. Suffixes set the field type; add _req to make any field required.
| Type | Suffix | Example | Required Form |
|---|---|---|---|
| Text (default) | — | {{First_Name}} | {{First_Name_req}} |
| Date | _dte | {{Due_Date_dte}} | {{Due_Date_dte_req}} |
| Email Address | _eml | {{Contact_Email_eml}} | {{Contact_Email_eml_req}} |
| Phone | _tel | {{Phone_tel}} | {{Phone_tel_req}} |
| Currency | _cur | {{Total_cur}} | {{Total_cur_req}} |
| URL | _url | {{Site_url}} | {{Site_url_req}} |
| Memo / Long Text | _mem | {{Notes_mem}} | {{Notes_mem_req}} |
| Number | _num | {{Quantity_num}} | {{Quantity_num_req}} |
| Decimal | _dec | {{Weight_dec}} | {{Weight_dec_req}} |
Repeating lists: group up to 4 columns per list name with dot notation, e.g. {{ItemNumber_req.product}}, {{ItemName_req.product}}, {{ItemQuantity.product}}, {{ItemCost.product}} — product is the list name shared across its columns.
The same REST endpoints that power the demo widgets on this page are what an LLM agent or your own tooling can call directly to generate workflows programmatically — the API contract for any generated form is downloadable from its dashboard entry.
One Spec, Three Generated Outputs
Genie Snap compiles a form UI, REST API, and document template from a single spec — file, text block, or prompt — usually in under 30 seconds. The same install also handles submission storage, task routing, and audit logging, so there's nothing extra to wire up.
🔒 No per-seat runtime fees, no hidden dependency chain: it's a single install that runs the same way online, on your LAN, or fully offline.
💡 Not sure where this fits? See 10 example workflow types it's used for
🔌 The demo widgets on this page call the same REST endpoints an LLM agent or your own tooling can call directly — nothing here is UI-only.
What's different
- Generating and previewing a workflow needs no account; saving it to your dashboard to activate it does require signing in.
- Editing the spec and regenerating takes seconds — there's no build/deploy cycle in between.
- Output is deterministic: the same text block compiles to the same form, API, and document every time, on every machine.
- Runs fully offline, on-prem, or air-gapped — no mandatory cloud dependency for the compiled runtime.
- The API's endpoint, auth model, and contract shape are identical across every generated workflow, and edited forms carry their own version stamp — so old integrations don't break when a form changes.
No Cloud Dependency
No GPU Requirement
No External Data Exposure
No Templates or Models
On-Premises & Air-Gapped AI Deployment
Genie Snap runs fully offline when deployed on-premises, including the optional local-LLM step.
- Optional Modules: Payment processing via Stripe is entirely optional. Email delivery requires standard SMTP or SendGrid configurations.
-
🧠 Intent Parsing Only: LLMs are completely optional and used strictly for generating text instructions from raw user intent. All subsequent application assets—forms, APIs, and document templates—are produced deterministically by the local framework.
- Local Model Orchestration: Ollama, Gemma-4, or other local open models can be installed entirely on-premises. This enables natural-language generation inside a completely closed, secure operational loop with zero phone-home packets.
- View the On-Premises Load Test Report (PDF) →
What the Platform Does
Eleven capabilities, from first request to signed document:
Product Guides & Video Walkthroughs
Explore detailed documentation or watch quick feature demonstrations to master your intake pipelines.
Technical Docs
Read detailed instructions to master form generation, custom APIs, validation rules, and offline execution setups.
eSignature
Route a single document — or several as one multi-document envelope — to multiple signers in order, or drop in any PDF/Word file and place signer fields directly on the page. REST API included. No third-party eSign vendor required.
Hosted Pricing
Self-Host Pricing
Air-Gapped ReadyLicensing is per database cluster, so you can license separate clusters per department, business unit, or environment. Each cluster runs independently with its own forms, submissions, and workflows.
- Submission limits are calculated on an annual cycle. Only new submissions created during the active license year count toward the annual limit.
- Historical submissions do not count toward the annual limit.
- If submissions are deleted, the available submission capacity for that license year increases by the same amount.
- At the start of each new license year, the submission counter resets, and the environment continues operating under the same license tier unless upgraded or changed.
Self-Host Download (Direct)
No industry-specific build required — it's used for anything with structured intake and a document at the end: compliance forms, KYC, patient intake, permit applications, supplier onboarding, and similar. The self-hosted / air-gapped option matters most where data can't leave the network (defense, healthcare, finance, government).