fix(http-request): accept any-case pair keys and return model-fixable errors

[ionut-mihalache-uipath]

Why

Agents using the HTTP Request tool failed at random with AGENT_RUNTIME.INVALID_INPUT_ARGUMENT:

headers.0.name  Field required [type=missing, input_value={'Name': 'Accept', 'Value': 'text/html'}]

The schema expects {"name", "value"} pairs, but models sometimes write {"Name", "Value"}, and pydantic matches keys case-sensitively. Because non-conversational agents have no node-level tool error handling, one malformed call out of many ended the whole job, and the model never got to retry.

What

• Pair keys in any case. A new _HttpRequestTool subclass lowercases name/value keys in headers and params before validation. Only the keys change; header and param values keep their case, so tokens and case-sensitive query values are untouched. _pairs_to_dict does the same lookup for input that skips validation.
• Model-fixable errors go back to the model. The tool now raises LangChain's ToolException and sets handle_tool_error=True, so these come back as ToolMessage(status="error") and the run continues:
  • schema mismatches that remain after the key fix (via an _invalid_input_error override)
  • a missing or non-string url
  • an unsupported method or a bad timeout
  • an SSRF-blocked host (the request is still blocked; only how the refusal is reported changes)
  • a host that doesn't resolve
  • timeouts and other httpx transport errors
• BaseUiPathStructuredTool._invalid_input_error has a wider return type (-> Exception). Other tools behave exactly as before.

Tests

• New tests cover mixed-case pair keys (values keep their case), schema mismatches returned to the model, read and connect errors, and a bad call through create_tool_node. That last one is the non-conversational graph path, which now yields an error tool message instead of raising.
• The existing "raises" tests now assert an error tool message.
• tests/agent: 2281 passed, 3 skipped. ruff, mypy and lint_httpx_client.py are clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_016T19pn6oVDUNv9YHzgQQUP

[Copilot]

Copilot review overview

🟢 Approval recommended

The implementation matches the stated behavior and includes focused coverage for the affected execution paths.

Review effort: Balanced
Findings: None

What changed in this PR

Improves HTTP Request tool resilience so agents can recover from casing mistakes and model-fixable failures.

Changes:

• Normalizes header and parameter pair keys case-insensitively.
• Returns validation, network, and SSRF errors as error ToolMessage responses.
• Adds coverage for normalization and non-conversational tool-node behavior.

File	Description
http_request_tool.py	Adds normalization and recoverable error handling.
base_uipath_structured_tool.py	Broadens validation-error return typing.
test_http_request_tool.py	Tests successful normalization and error responses.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
97.5% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud