Skip to main content

​
Install

The CLI is included when you install SuperDialog: 
pip install superdialog
superdialog --help

​
superdialog chat

Interactive terminal chat against a flow. No infrastructure, no API keys for Unpod, no phone number. The fastest way to test your dialog machine exactly as an end user would experience it. 
superdialog chat kyc.json

> Hello, I need to verify my KYC.
Agent: Sure! Could you please provide the last 4 digits of your Aadhaar?
> 1234
Agent: Thank you. Could you also confirm your date of birth?
> 1990-05-15
Agent: Perfect, your KYC has been verified. Have a great day!
Options: 
superdialog chat kyc.json --llm anthropic/claude-haiku-4-5     # override runtime model
superdialog chat kyc.json --traversal-dir ./history             # save session JSON on completion
superdialog chat kyc.json --adapter toolcall                    # "toolcall" (default) or "llm"
FlagDefaultDescription
--flowflow.jsonPath to flow file (JSON or YAML)
--llmopenai/gpt-4o-miniModel URI for runtime
--adaptertoolcalltoolcall (1 LLM call/turn) or llm (2 calls/turn, useful for debugging)
When to use: during initial flow design, prompt tuning, and eval dataset collection - before any voice infrastructure is involved.

​
superdialog flow lint

Validate a flow’s graph structure. Checks for unreachable nodes, missing edges, undefined slots, and other structural errors. 
superdialog flow lint kyc.json

✅ kyc.json - no errors found (7 nodes, 9 edges)
When to use: after hand-editing a flow JSON, or as a CI check before deployment.

​
superdialog flow draw

Render a Mermaid diagram of the flow graph. Paste the output into any Mermaid renderer. 
superdialog flow draw kyc.json

graph TD
  start --> collect_aadhaar
  collect_aadhaar --> verify_aadhaar
  verify_aadhaar -- verified --> collect_dob
  verify_aadhaar -- failed --> retry
  retry --> collect_aadhaar
  collect_dob --> done
When to use: to visualise and share flow structure, in design reviews, or when debugging unexpected state transitions.

​
superdialog flow generate

Bootstrap a flow.json from a plain-language prompt. Equivalent to calling create_dialog_flow(prompt=..., llm=...) from Python. 
superdialog flow generate "Confirm KYC. Ask for Aadhaar last 4. Confirm DOB." \
  --llm openai/gpt-5.1 \
  --output kyc.json
Options:
FlagDefaultDescription
--llmopenai/gpt-4o-miniModel URI used to generate the flow
--outputflow.jsonOutput file path
--from-Read prompt from a file instead of inline
# Prompt from a file
superdialog flow generate --from prompt.txt --llm openai/gpt-5.1 --output kyc.json
When to use: to quickly prototype a new flow, then refine the JSON by hand or re-generate with a better prompt.

​
superdialog eval (v0.3 - planned)

Run an eval harness against a corpus of test cases. 
superdialog eval kyc.json tests/kyc-corpus.jsonl \
  --llm openai/gpt-5.1 \
  --llm anthropic/claude-haiku-4-5 \
  --metrics accuracy latency_p95
Outputs a per-LLM comparison report.

​
Traversal history

Any command that runs conversations supports --traversal-dir. When set, a timestamped JSON file is written for each completed session capturing:
  • Every node visited (in order)
  • Every turn (user text + agent reply)
  • All collected slot values
superdialog chat kyc.json --traversal-dir ./traversal_history
Use these files to build eval corpora, debug unexpected flow paths, and audit production conversations.