fix(llm): image detail field + /v1 base URL normalization (#2380)

* fix(llm): add image detail field, auto-append /v1 to base URL (#2378, #1934)

Set detail: "auto" on ImageUrl construction so providers requiring the
field (e.g. MiniMax) no longer reject vision requests. Normalize
OpenAI-compatible base URLs by appending /v1 when missing, fixing 404s
for local model servers (MLX, vLLM, llama.cpp) using bare URLs.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(llm): scope /v1 normalization to bare host-only URLs

Address review feedback:

- Only append /v1 when the URL has no path component (bare
  scheme://host[:port]). URLs with existing paths like Zai's
  /api/paas/v4 or Gemini's /v1beta/openai are now left unchanged.
- Use case-insensitive check for /v1 suffix to prevent double-suffixing
  URLs like http://localhost:8080/V1.
- Document why Ollama is intentionally excluded from normalization
  (uses /api/chat, not /v1/chat/completions).
- Add test cases for real provider URLs from providers.json (Zai,
  Gemini) and case-insensitive /V1.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* chore: update gimli 0.33.1 -> 0.33.0 (yanked crate)

gimli v0.33.1 was yanked on crates.io, causing cargo-deny to fail.
Downgrade to v0.33.0 which is the latest non-yanked release compatible
with wasmtime 43.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: serrrfirat <f@nuff.tech>

Cargo.lock

@@ -3112,9 +3112,9 @@ dependencies = [
 
 [[package]]
 name = "gimli"
-version = "0.33.1"
+version = "0.33.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "19e16c5073773ccf057c282be832a59ee53ef5ff98db3aeff7f8314f52ffc196"
+checksum = "0bf7f043f89559805f8c7cacc432749b2fa0d0a0a9ee46ce47164ed5ba7f126c"
 dependencies = [
  "fnv",
  "hashbrown 0.16.1",

src/agent/attachments.rs

@@ -43,7 +43,7 @@ pub fn augment_with_attachments(
             image_parts.push(ContentPart::ImageUrl {
                 image_url: ImageUrl {
                     url: data_url,
-                    detail: None,
+                    detail: Some("auto".to_string()),
                 },
             });
         }
@@ -242,6 +242,26 @@ mod tests {
         }
     }
 
+    #[test]
+    fn image_url_includes_detail_auto() {
+        let mut att = make_attachment(AttachmentKind::Image);
+        att.mime_type = "image/png".to_string();
+        att.data = vec![0x89, 0x50, 0x4E, 0x47]; // fake PNG header
+
+        let result = augment_with_attachments("check", &[att]).unwrap();
+        assert_eq!(result.image_parts.len(), 1);
+        match &result.image_parts[0] {
+            ContentPart::ImageUrl { image_url } => {
+                assert_eq!(
+                    image_url.detail.as_deref(),
+                    Some("auto"),
+                    "detail field must be set to 'auto' for provider compatibility"
+                );
+            }
+            other => panic!("Expected ImageUrl, got: {:?}", other),
+        }
+    }
+
     #[test]
     fn document_with_extracted_text() {
         let mut att = make_attachment(AttachmentKind::Document);

src/llm/mod.rs

@@ -285,7 +285,8 @@ fn create_openai_compat_from_registry(
 
     let mut builder = openai::Client::builder().api_key(&api_key);
     if !config.base_url.is_empty() {
-        builder = builder.base_url(&config.base_url);
+        let base_url = normalize_openai_base_url(&config.base_url);
+        builder = builder.base_url(&base_url);
     }
     if !extra_headers.is_empty() {
         builder = builder.http_headers(extra_headers);
@@ -707,6 +708,34 @@ pub fn create_gemini_oauth_provider(config: &LlmConfig) -> Result<Arc<dyn LlmPro
     Ok(Arc::new(provider))
 }
 
+/// Normalize an OpenAI-compatible base URL by appending `/v1` when the URL
+/// contains no path (bare `scheme://host[:port]`).
+///
+/// rig-core's `openai::Client` does not auto-append `/v1/` to the base URL,
+/// so local model servers (MLX, vLLM, llama.cpp) using bare URLs like
+/// `http://localhost:8080` get 404s. This mirrors the old
+/// `NearAiChatProvider::api_url()` behavior.
+///
+/// URLs that already carry a path — including non-`/v1` versioned paths such
+/// as Zai's `/api/paas/v4` or Gemini's `/v1beta/openai` — are returned
+/// unchanged so we don't corrupt provider-specific endpoints.
+///
+/// **Note:** This is intentionally applied only to `OpenAiCompletions`-protocol
+/// providers. Ollama uses `/api/chat` (not `/v1/chat/completions`) and its
+/// rig-core client handles the path internally, so normalization is not needed.
+fn normalize_openai_base_url(url: &str) -> String {
+    let trimmed = url.trim_end_matches('/');
+    if trimmed.to_ascii_lowercase().ends_with("/v1") {
+        return trimmed.to_string();
+    }
+    match url::Url::parse(trimmed) {
+        Ok(parsed) if parsed.path().is_empty() || parsed.path() == "/" => {
+            format!("{trimmed}/v1")
+        }
+        _ => trimmed.to_string(),
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -867,4 +896,59 @@ mod tests {
         let config = test_llm_config();
         assert_eq!(config.cheap_model_name(), None);
     }
+
+    #[test]
+    fn test_normalize_openai_base_url_appends_v1_for_bare_hosts() {
+        assert_eq!(
+            normalize_openai_base_url("http://localhost:8080"),
+            "http://localhost:8080/v1"
+        );
+        assert_eq!(
+            normalize_openai_base_url("http://localhost:8080/"),
+            "http://localhost:8080/v1"
+        );
+        assert_eq!(
+            normalize_openai_base_url("https://my-server.example.com"),
+            "https://my-server.example.com/v1"
+        );
+    }
+
+    #[test]
+    fn test_normalize_openai_base_url_leaves_v1_alone() {
+        assert_eq!(
+            normalize_openai_base_url("http://localhost:8080/v1"),
+            "http://localhost:8080/v1"
+        );
+        assert_eq!(
+            normalize_openai_base_url("http://localhost:8080/v1/"),
+            "http://localhost:8080/v1"
+        );
+        assert_eq!(
+            normalize_openai_base_url("https://api.openai.com/v1"),
+            "https://api.openai.com/v1"
+        );
+        // Case-insensitive: /V1 should not get double-suffixed
+        assert_eq!(
+            normalize_openai_base_url("http://localhost:8080/V1"),
+            "http://localhost:8080/V1"
+        );
+    }
+
+    #[test]
+    fn test_normalize_openai_base_url_preserves_existing_paths() {
+        // Non-/v1 versioned paths from real providers must stay unchanged
+        assert_eq!(
+            normalize_openai_base_url("https://api.z.ai/api/paas/v4"),
+            "https://api.z.ai/api/paas/v4"
+        );
+        assert_eq!(
+            normalize_openai_base_url("https://generativelanguage.googleapis.com/v1beta/openai"),
+            "https://generativelanguage.googleapis.com/v1beta/openai"
+        );
+        // Custom subpaths should also stay unchanged
+        assert_eq!(
+            normalize_openai_base_url("https://api.example.com/custom"),
+            "https://api.example.com/custom"
+        );
+    }
 }