Local gateway + Ollama code example broken — base_url missing /v1 — Epic Games

[anshulbasia27]

Summary

Fix broken code example in the Ollama local setup docs. The base_url parameter was missing the required /v1 path suffix, causing SDK errors when connecting to a local Portkey gateway.

What changed

• integrations/llms/ollama.mdx:
  • Fixed base_url from http://localhost:8787 to http://localhost:8787/v1 in the Local Setup Python example
  • Added a Node.js code example alongside the Python one (wrapped in <CodeGroup>)
  • Clarified the <Note> block at the bottom of the page to distinguish between custom_host (Ollama URL, no /v1) and base_url (gateway URL, requires /v1)

Customer context

Reported by Epic Games. The customer provided screenshots showing:

1. Working response when base_url includes /v1 (successful chat completion via Ollama)
2. Error when /v1 is omitted

Acceptance criteria

• base_url in Local Setup example includes /v1
• Node.js example added for parity
• <Note> block clarified to avoid confusion between custom_host and base_url

[anshulbasia27]

@Cursor-AI Please review this PR.

Context

This PR was automatically generated by Altor to address a customer-reported product gap for Epic Games.

Field	Value
Customer	Epic Games
Category	support-gap
Gap	Local gateway + Ollama code example broken — base_url missing /v1

Requirement

Local gateway + Ollama code example broken — base_url missing /v1

Acceptance Criteria

Please verify each of the following is satisfied by the implementation:

• Verify the fix resolves the customer-reported issue without regressions

References

• Pylon ticket #0: https://app.usepylon.com/support/issues/views/all-issues?issueNumber=0

Review Focus

• Correctness: Does the implementation fully satisfy the requirement and acceptance criteria above?
• Edge cases: Are there scenarios involving Local gateway + Ollama code example broken — base_url missing /v1 that are not handled?
• Regressions: Could any existing behaviour be broken by these changes?
• Code quality: Are there obvious improvements to readability, error handling, or performance?

Please call out anything that looks incomplete or incorrect before this is merged.

[akhilmmenon]

Reviewed integrations/llms/ollama.mdx against the issue context in [#812]
Looks good and meets the stated acceptance criteria:
Local setup now correctly uses base_url / baseURL with /v1 (http://localhost:8787/v1), which fixes the reported local gateway failure.
Node.js example is present alongside Python for parity.

The note clearly distinguishes:
custom_host / customHost (Ollama URL, without /v1)
base_url / baseURL (local Portkey gateway URL, with /v1)

I don’t see regressions in the updated examples.
Non-blocking note: runtime execution wasn’t re-verified in this review pass; this is a content/accuracy validation.