v0.0.6

Added

• Authentication and authorization layer to use the Rest API.
• /quotas endpoint to fetch the service usage

Fixed

• CORS: validation of preflight calls.
• Response's content-type header value.

Changed

• OpenAI's tokens consumption tracking.

v0.0.6beta

Added

• Rest API interface to the core logic. Find the specification here.

Fixed

• OpenAI's prediction parsing and deserialization when the graph's JSON is surrounded by text.

Example

GIVEN: The OpenAI chat response's content below

Here's the C4 diagram for a Python web server reading from an external Postgres database:

{
  "title": "Python Web Server Reading from External Postgres Database",
  "nodes": [
    {"id": "0", "label": "Web Server", "technology": "Python"},
    {"id": "1", "label": "Postgres", "technology": "Postgres", "external": true, "is_database": true}
  ],
  "links": [
    {"from": "0", "to": "1", "label": "reads from Postgres", "technology": "TCP"}
  ],
  "footer": "C4 Model"
}

The diagram shows two nodes: a Python web server and an external Postgres database. The web server reads data from the Postgres database over TCP.

WHEN: apply new deserialization and parsing logic

THEN: get the following graph definition:

{"title":"Python Web Server Reading from External Postgres Database","nodes":[{"id":"0","label":"Web Server","technology":"Python"},{"id":"1","label":"Postgres","technology":"Postgres","external":true,"is_database":true}],"links":[{"from":"0","to":"1","label":"reads from Postgres","technology":"TCP"}],"footer":"C4 Model"}

v0.0.5

Added

• User input validation according to the user status, i.e. registered/non-registered split
• Improved user feedback in case the model fails to predict a graph
• Trigger diagram generation process on the ctrl+enter keys press
• Cancellation of requests over 10 sec. on the escape key press
• Hide popup on the escape key press

Changed

• Refactored the webclient codebase:
  • typescript
  • vite
  • tests coverage improvement

v0.0.4

Added

• Support for the User C4 macro.
• Diagram legend by default. It can be removed upon an explicit prompt's request only.

Examples:

The prompt yields the diagram with the legend:

diagram of three boxes 

The prompts yield the diagram without the legend:

diagram of three boxes without legend

diagram of three boxes, no legend

diagram of three boxes with no legend

diagram of three boxes. remove legend

Fixed

• (#56) Moved from the model from "code-davinci-002" to "gpt-3.5-turbo" following on the OpenAI announcement to discontinue support for the Codex API.

v0.0.3

Fixed

• Migrated the core logic to GCP to bypass the OpenAI latency exceeding the AWS GW request limit of 29 sec. Note that
  the tf codebase for AWS infra is preserved for the release.

Changed

• Refactored the core logic:
  • Port-adaptor
    approach
  • Split sub-packages to dedicated modules in the pkg directory. Note that the packages below released as v0.0.1:
    • HTTP client with the backoff-retry mechanism
    • GCP Secretsmanager
    • Postgres client
    • OpenAI client
• tf state with the Neon db provisioning migrated to GCP.

v0.0.2

Changed

• Model configurations were changed:

{
  "model": "code-davinci-002",
  "max_tokens": 768,
  "temperature": 0.2,
  "best_of": 3,
  "top_p": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "stop": ["\n"]
}

• (minor) Added the logo

Security

• Added the vault to manage access credentials to improve security

v0.0.1

Beta release: the announcement.

Added

• Core server-side logic to deliver C4 model diagrams
  • OpenAI agent
  • PlantUML agent
  • AWS Lambda interface
  • HTTP interface
• Webclient:
  • Freetext intput with the input validation
  • Output
  • Go and download buttons

Open IA Model

Configuration:

{
  "model": "code-cushman-001",
  "max_tokens": 768,
  "temperature": 0.2,
  "best_of": 3,
  "top_p": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "stop": ["\n"]
}