We need to import all tools so that they are available for deserialization in webhooks.

Infer and verify the repository from issue content. Raises: RepositoryNotFoundError: If no valid repository can be determined.

Inherit configuration from parent conversation if not explicitly provided. This ensures sub-conversations automatically inherit: - Sa

Inject flat top-level convenience fields for the SDK. The SDK's ``get_llm()`` and ``get_mcp_config()`` read ``llm_model``, ``llm_api_key``, `

Check if a legacy value is encrypted by trying to decrypt it

Check if PR was closed (regardless of conversation URL)

Check if the query is a repository URL.

Lift nested api keys to ``llm_api_key`` and mask the JSON diff.

Query AWS Bedrock for available foundation models. This is a synchronous boto3 call; callers should run it via ``call_sync_from_async

Load workspace skills onto a conversation request's agent. Used by both the LLM and ACP arms of ``_build_start_conversation_request_f

Build a real ``ConversationInfo`` with a real ``Agent``. Using real Pydantic models (rather than ``MagicMock(spec=...)``) exercises the discr

Whether to probe packages and runtime versions into the manifest.

Best-effort removal of the OpenHands webhook from Jira during teardown. Symmetric with :func:`_maybe_register_webhook`. When an admin PAT is

Return the managed LLM key every member row should carry, if any.

Normalize nested LLM settings inside ``agent_settings_diff``.

Ensure automation parsing can identify the Bitbucket DC event key. Bitbucket DC sends the event key both in ``X-Event-Key`` and, for normal d

Monotonic clock; isolated for tests.

Return a JSON OAuth-style error response.

Parse a comment_created event.

Parse action payload and extract message_ts, thread_ts, and selected value. This handles the different payload structures for button clicks v

Parse the raw ``INSTANCE_DEFAULT_MARKETPLACES`` value into raw dicts. Two formats are supported. JSON is tried on the *whole* value first so that

Parse an issue_updated event for label changes.

Parse ``npm ls --json`` output (top-level deps) into ``{name: version}``.

Parse ``pip list --format=json`` output into ``{name: version}``.

Parse the `key=value` runtime lines into ``{python, node, os}``.

Parse YAML frontmatter from a skill markdown file. Returns the frontmatter dict, or None if parsing fails.

Backfill the seeded ``Default`` profile onto ``org.llm_profiles``. Runs once per org during the pre-llm_profiles → llm_profiles upgrade:

Persist organization with compensation on failure. If database persistence fails, cleans up LiteLLM resources. Args:

Poll the agent server for the conversation title. Args: httpx_client: The HTTP client to use for requests. url: The conversation

Post a resolver response message to the configured Slack channel.

Strip response-only LLM fields without changing provider names.

Collect installed package and runtime versions.

Process commits from a single GraphQL page

Handle Jira DC webhook events.

Process message and return view if authenticated, or raise SlackError. Returns: SlackViewInterface if user is authenticated and r

Process pending messages queued before conversation was ready. Messages are delivered concurrently to the agent server. After processing,

Process PR comments from a single GraphQL page

Convert raw Bitbucket DC comment dicts to Comment objects.

Convert raw Bitbucket Cloud comment dicts to :class:`Comment`s. Sorted by ``created_on`` ascending; only the most recent ``max_commen

Process reviews and review comments from a single GraphQL page

Create a select query that returns public conversations with SAAS metadata. This joins with conversation_metadata_saas to retrieve the user_i

Pull Docker image with periodic progress logging every 5 seconds.

Build a simple JSON response that includes the details of the rate limit that was hit.

Return the response body as JSON, falling back to text. Infrastructure errors (e.g. 502/503 from a load balancer) often return a non-

Replace the interactive repo selector with a non-interactive status.

Return the agent's final response without asking the LLM to summarize.

Return the agent's final response without asking the LLM to summarize.

Ask the agent to produce a summary of its work and return the agent response.

Ask the agent to produce a summary of its work and return the agent response. NOTE: This method now returns a string (the agent server's resp