Generate a scope
What the model returns, how long generation takes, and what Byteflare does when a response comes back malformed.
Who this is for: Anyone generating or regenerating a scope, and anyone diagnosing a bad result.
What a generated scope contains
- An executive summary.
- Sections, each with a title, a description, and a list of deliverables.
- A pricing table of line items with amounts and optional notes.
- A timeline.
- Assumptions, exclusions and client responsibilities.
- Payment milestones.
- Optional add-ons with prices.
How long it takes
Scope generation streams progress while it runs and is allowed up to two minutes. If the model is overloaded, Byteflare retries with an increasing delay before giving up. A request that genuinely times out returns a retriable error rather than a partial scope.
What happens when the model returns something malformed
Byteflare does not simply fail. The raw response goes through a layered recovery pipeline before anything is saved.
Scope recovery pipeline
- ExtractPull the JSON object out of the model's response.
- NormaliseCoerce near-miss shapes into the expected structure.
- ValidateCheck the result against the scope schema.
- Merge defaultsFill in empty assumptions, exclusions, responsibilities, milestones and add-ons, then validate again.
- RepairFix structural problems and validate a third time.
- Ask the model to correct itselfSend the failed output and the validation errors back for correction, up to two attempts.
- Synthesise a stubAs a last resort, build a minimal scope from the brief so you have something to edit rather than an error screen.
Usage limits
On the free plan, AI generation is capped at 20 requests per day and 5 per hour, counted across both question generation and scope generation. Pro, trial and Team accounts are not subject to that daily or hourly cap.
Separately, every account has a short-window request rate limit that protects the service from bursts. If you hit it, the product asks you to wait a moment and try again. It resets on its own.
Which model runs
Free accounts and accounts with Pro access use different scope prompts and different models. Byteflare uses Anthropic's Claude models as the primary provider and Amazon Bedrock as a fallback when the primary provider is unavailable. Both receive the same request content. Privacy, data and AI covers what that content is.