dsh-custom-headers

In real-world deployments, model requests rarely go straight to the provider. They pass through enterprise gateways, custom proxies, or OpenAI-compatible endpoints that demand extra headers—tenant IDs, billing tags, custom auth schemes—often different per model. DSH's provider profile supports a headers field, but it applies to the entire route, making it impossible to target individual models without affecting every one under that route.

dsh-custom-headers fills that gap at model granularity. Define any number of named header profiles in the plugin settings panel, each a set of key-value pairs; then pick one per model from a dropdown in the model directory. On every call the plugin resolves the selected profile into request headers and injects them before the pi-ai adapter dispatches. Change a profile or switch the selection and the next request picks it up immediately—no restart required. All data lives in the DSH user-settings document; deselecting or deleting a profile restores the original state with no residue.

Who benefits: teams routing through enterprise gateways for tenant and cost-center billing; developers hitting non-standard OpenAI-compatible endpoints (self-hosted vLLM, one-api, various relays) that require extra identification headers; and engineers tagging specific model traffic for canary releases or observability. The plugin is MIT-licensed, introduces no external storage, and works exclusively with the pi-ai adapter path.