fix(instrumentation-py): repair Agno, Aleph Alpha, and Anthropic OTel 2.x spans

[Nightingalelyy]

Summary

• preserve the enclosing workflow name across Agno agent, team, chat, tool, and event spans, and retain tool calls in structured entity output
• classify Aleph Alpha evaluate and explain as non-generative task spans without fabricated model usage, while retaining structured results and chat tool-call output
• preserve Anthropic provider HTTP status codes and active workflow names on chat, tool, and error spans
• add focused regressions and patch release intent for all three packages

Why.

The OTel compatibility audit found that these packages split logical workflows, misclassified Aleph Alpha analytical operations, and flattened real Anthropic provider failures to generic 500 records.

Root cause.

• native Agno spans replaced or omitted the active workflow context
• Aleph Alpha treated every non-chat/non-embedding method as a completion
• Anthropic error export ignored status fields carried by SDK exceptions, and injected spans did not read the active workflow context

Validation

• focused tests: Agno 14 passed / 1 opt-in skipped; Aleph Alpha 7 passed; Anthropic 15 passed

• all three wheels built, installed together in a clean Python 3.11 environment, and imported from the wheel artifacts

• all 15 corresponding examples passed against locally linked packages under marker otel2-fix-py-group-08-20260815T201000Z

• Respan MCP exact-marker audit: 15/15 trace trees and 51/51 full records inspected; exact span counts, connected hierarchy, workflow grouping, content, usage, tools, and expected 404 status passed

• release inventory, release-intent, missing-intent, and git diff --check validation passed

• Contribution re-audit against every current document under contribution/ passed after 77f1b68: Aleph Alpha text completion uses canonical chat request type; Anthropic automatic span-kind and deprecated tool aliases are absent; definitions/calls and tool execution I/O use canonical JSON.

• Fresh correction validation: Agno 14 passed / 1 skipped, Aleph Alpha 7 passed, Anthropic 15 passed; corrected-package wheels, release inventory/intents, formatting/import, and diff checks passed; marker otel2-pr-contract-corrections-20260818T133300Z produced 2 representative traces / 6 full records for Aleph Alpha and Anthropic, all inspected.

External follow-ups

• The platform currently retains Agno/Aleph canonical tool attributes and structured message content but does not project them into its top-level tool arrays; this is recorded as a backend follow-up rather than addressed with prohibited compatibility aliases.

Companion examples

• fix(examples-py): validate Agno, Aleph Alpha, and Anthropic OTel 2.x spans respan-example-projects#55

[greptile-apps]

Your trial has ended. Reactivate Greptile to resume code reviews.