telegras 0.3.0

Telegram + FastAPI ingestion core with pluggable publication backends

telegras

Standalone Telegram ingestion core extracted from tg-wp-bridge.

Parsed Telegram API

The generated Telegram Bot API models live in tg_api_parsed. They are copied from the reviewed telegram-api-extract artifacts and expose only the types used by telegras:

• tg_api_parsed.Update
• tg_api_parsed.WebhookInfo
• tg_api_parsed.SetWebhookRequest
• tg_api_parsed.SendMessageRequest
• tg_api_parsed.GetMeResponse

CI & Releases

• CI: ci.yml runs uv run pytest -q and uv run --extra docs mkdocs build --strict on pushes/PRs against main.
• Releases: publish.yml triggers on push tags v*, builds the package, checks artifacts, and publishes to PyPI. Configure a PyPI trusted publisher (OIDC) so secrets.PYPI_TOKEN can stay empty while GitHub forwards identity securely. If you prefer the classic API token, create a repository secret named PYPI_TOKEN.

Local development

• Install: uv pip install -e .[dev]
• Run API: uv run uvicorn telegras.app:app --reload
• Run tests: uv run pytest

Bot mode

Set BOT_MODE to control update transport:

• webhook (default)
• polling

Examples:

• BOT_MODE=webhook uv run uvicorn telegras.app:app --reload
• BOT_MODE=polling uv run uvicorn telegras.app:app --reload

API submodule

telegras now includes a curated Telegram API runtime subset in telegras.api:

• telegras.api.client.TelegramBotAPI
• telegras.api.getting_updates (Update, WebhookInfo, SetWebhookRequest)
• telegras.api.methods (SendMessageRequest, GetMeResponse)

This subset is extracted from reviewed generated specs in docs/telegram_api/ and used by telegras.telegram_api.

Webhook Attachments

Webhook attachments are composable match rules bound to handler identifiers.

Attachment schema

{
  "name": "blog-channel",
  "handler": "handlers.python:eval",
  "handler_args": ["result = '{{ message.title }}'"],
  "enabled": true,
  "priority": 10,
  "stop_on_match": false,
  "when": {
    "op": "all",
    "children": [
      {
        "op": "leaf",
        "leaf": {
          "field": "chat.type",
          "match": "eq",
          "value": "channel"
        }
      },
      {
        "op": "leaf",
        "leaf": {
          "field": "message.text",
          "match": "regex",
          "value": "#blog\\b"
        }
      }
    ]
  }
}

API endpoints

• GET /v1/webhook-attachments
• POST /v1/webhook-attachments
• DELETE /v1/webhook-attachments/{name}
• POST /v1/webhook-attachments/match
• POST /v1/webhook-attachments/execute

Protected introspection API

Set INTROSPECTION_TOKEN and call with Authorization: Bearer <token>.

• GET /internal/introspection/config
• GET /internal/introspection/attachments
• POST /internal/introspection/attachments
• PUT /internal/introspection/attachments/{name}
• DELETE /internal/introspection/attachments/{name}
• GET /internal/introspection/webhook-executions
• GET /internal/introspection/handlers
• GET /internal/introspection/match-criteria

Exact examples

Register a shell listing handler:

{
  "name": "list-temp",
  "handler": "handlers.shell:ls",
  "handler_args": ["/home", "/tmp"],
  "when": {
    "op": "leaf",
    "leaf": {
      "field": "chat.type",
      "match": "eq",
      "value": "group"
    }
  }
}

Register a Python eval handler with parse-result templates:

{
  "name": "mail-on-title",
  "handler": "handlers.python:eval",
  "handler_args": [
    "import mails",
    "mails.send_mail('{{ message.title }}')",
    "log.info(f\"full message: {{ message.full }}\")"
  ],
  "when": {
    "op": "all",
    "children": [
      {
        "op": "leaf",
        "leaf": {
          "field": "chat.type",
          "match": "eq",
          "value": "group"
        }
      },
      {
        "op": "leaf",
        "leaf": {
          "field": "message.text",
          "match": "contains",
          "value": "urgent"
        }
      }
    ]
  }
}

Register a grouped handler chain with parser extraction:

{
  "name": "grouped-mail",
  "handler_chain": [
    {
      "handler": "handlers.python:eval",
      "handler_args": ["title = '{{ match.title }}'"]
    },
    {
      "handler": "handlers.python:eval",
      "handler_args": ["log.info('full=' + '{{ message.full }}')"]
    }
  ],
  "execution_mode": "sequential",
  "stop_on_error": true,
  "parse": {
    "regex": {
      "title": "^(?P<title>[^\\n]+)"
    },
    "parser_ref": null,
    "allow_partial": true
  },
  "parse_mode": "warn",
  "when": {
    "op": "leaf",
    "leaf": {
      "field": "chat.type",
      "match": "eq",
      "value": "group"
    }
  }
}

Execute matching attachments for an incoming update:

{
  "update": {
    "update_id": 10001,
    "message": {
      "message_id": 33,
      "date": 1700000000,
      "chat": {
        "id": 444,
        "type": "group",
        "title": "Ops"
      },
      "text": "Status report\nEverything looks green."
    }
  }
}

Built-in handlers

• handlers.shell:ls
  • Args: path strings
  • Result: map of path to directory entries (or error)
• handlers.shell:sh
  • Args: shell command fragments, joined with &&
  • Result: returncode, stdout, stderr, command
• handlers.python:eval
  • Args: Python snippets executed in order
  • Available variables: message, chat, update, log
  • Result: serializable locals snapshot

For the complete user guide to matcher expressions and handler execution, see:

• docs/webhook-attachments.md

Template fields for handler_args

{{ ... }} placeholders are supported:

• {{ message.title }} first line of text/caption
• {{ message.full }} full text/caption
• {{ message.text }} full text/caption
• {{ message.has_media }} boolean
• {{ message.media_type }} detected media kind
• {{ chat.id }} chat id
• {{ chat.type }} chat type
• {{ chat.title }} chat title
• {{ update.update_id }} Telegram update id
• {{ update.kind }} detected update kind
• {{ match.<key> }} parser output values

Telegram API extraction tooling

Extraction and generation scripts that parse Telegram HTML API docs were moved to:

• docs/telegram_api_extraction/

Purpose and script-level documentation:

• docs/telegram_api_extraction/README.md

Backward-compatible wrappers remain in docs/*.py.