Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

fix: support arbitrary version segments in openai-compat base URLs (#6)

Lucas Oliveira87714da417c3
diff --git a/crates/ai-memory-llm/src/embedding.rs b/crates/ai-memory-llm/src/embedding.rs
index 2e9df444..8607badb 100644
--- a/crates/ai-memory-llm/src/embedding.rs
+++ b/crates/ai-memory-llm/src/embedding.rs
@@ -114,7 +114,7 @@ impl Embedder for OpenAiEmbedder {
}
async fn embed(&self, text: &str) -> LlmResult<Vec<f32>> {
- let url = format!("{}/v1/embeddings", normalize_openai_base(&self.base_url));
+ let url = normalize_openai_base(&self.base_url, "embeddings");
debug!(url, model = %self.model, "POST openai/embeddings");
let req = OpenAiEmbeddingRequest {
input: text,
@@ -222,7 +222,7 @@ impl Embedder for VoyageEmbedder {
}
async fn embed(&self, text: &str) -> LlmResult<Vec<f32>> {
- let url = format!("{}/v1/embeddings", normalize_openai_base(&self.base_url));
+ let url = normalize_openai_base(&self.base_url, "embeddings");
let req = VoyageRequest {
input: [text],
model: &self.model,
diff --git a/crates/ai-memory-llm/src/openai.rs b/crates/ai-memory-llm/src/openai.rs
index 842ce38c..5cc54b08 100644
--- a/crates/ai-memory-llm/src/openai.rs
+++ b/crates/ai-memory-llm/src/openai.rs
@@ -15,19 +15,35 @@ use crate::types::{ChatRequest, ChatResponse, Role, Usage};
/// Default OpenAI API base.
pub const DEFAULT_BASE_URL: &str = "https://api.openai.com";
-/// Trim a trailing `/` and one optional `/v1` segment off an
-/// OpenAI-style base URL so the caller can append `/v1/<endpoint>`
-/// safely. Tolerates both conventions found in the wild:
+/// Build the full URL for an OpenAI-style endpoint. Tolerates the
+/// conventions found in the wild:
/// * `https://api.openai.com` (OpenAI's own docs)
/// * `https://openrouter.ai/api/v1` (OpenRouter's docs)
/// * `http://localhost:11434/v1` (Ollama's openai-compat path)
+/// * `https://api.z.ai/api/coding/paas/v4` (Z.AI)
///
/// Without this, half the providers produce `…/v1/v1/…` 404s the
/// first time consolidation runs.
#[must_use]
-pub fn normalize_openai_base(url: &str) -> String {
- let s = url.trim_end_matches('/');
- s.strip_suffix("/v1").unwrap_or(s).to_string()
+pub fn normalize_openai_base(base: &str, endpoint: &str) -> String {
+ let s = base.trim_end_matches('/');
+
+ if s.ends_with(&format!("/{endpoint}")) {
+ return s.to_string();
+ }
+
+ if last_segment_is_version(s) {
+ return format!("{s}/{endpoint}");
+ }
+
+ format!("{s}/v1/{endpoint}")
+}
+
+fn last_segment_is_version(url: &str) -> bool {
+ url.split('/').next_back().is_some_and(|seg| {
+ let digits = seg.strip_prefix('v').unwrap_or("");
+ !digits.is_empty() && digits.len() <= 2 && digits.chars().all(|c| c.is_ascii_digit())
+ })
}
/// OpenAI Chat Completions-backed provider.
@@ -211,10 +227,7 @@ impl OpenAiProvider {
}
async fn post<B: Serialize>(&self, body: &B) -> LlmResult<OpenAiResponse> {
- let url = format!(
- "{}/v1/chat/completions",
- normalize_openai_base(&self.base_url)
- );
+ let url = normalize_openai_base(&self.base_url, "chat/completions");
debug!(url, "POST openai");
let resp = self
.client
@@ -249,32 +262,69 @@ mod tests {
use super::normalize_openai_base;
#[test]
- fn normalize_strips_trailing_slash_and_v1() {
- // The three conventions found in the wild.
+ fn normalize_openai_base_chat_completions() {
+ let ep = "chat/completions";
+
+ assert_eq!(
+ normalize_openai_base("https://api.openai.com", ep),
+ "https://api.openai.com/v1/chat/completions"
+ );
+ assert_eq!(
+ normalize_openai_base("https://api.openai.com/", ep),
+ "https://api.openai.com/v1/chat/completions"
+ );
+ assert_eq!(
+ normalize_openai_base("https://openrouter.ai/api/v1", ep),
+ "https://openrouter.ai/api/v1/chat/completions"
+ );
+ assert_eq!(
+ normalize_openai_base("http://localhost:11434/v1", ep),
+ "http://localhost:11434/v1/chat/completions"
+ );
+ // /v123 must not be treated as a version segment.
+ assert_eq!(
+ normalize_openai_base("https://example.com/v123", ep),
+ "https://example.com/v123/v1/chat/completions"
+ );
+ // Z.AI-style: non-v1 version segment in the path.
+ assert_eq!(
+ normalize_openai_base("https://api.z.ai/api/coding/paas/v4", ep),
+ "https://api.z.ai/api/coding/paas/v4/chat/completions"
+ );
+ // Full endpoint URL already provided (Z.AI or GitHub Copilot style).
+ assert_eq!(
+ normalize_openai_base("https://api.z.ai/api/coding/paas/v4/chat/completions", ep),
+ "https://api.z.ai/api/coding/paas/v4/chat/completions"
+ );
+ assert_eq!(
+ normalize_openai_base("https://api.githubcopilot.com/chat/completions", ep),
+ "https://api.githubcopilot.com/chat/completions"
+ );
+ }
+
+ #[test]
+ fn normalize_openai_base_embeddings() {
+ let ep = "embeddings";
+
assert_eq!(
- normalize_openai_base("https://api.openai.com"),
- "https://api.openai.com"
+ normalize_openai_base("https://api.openai.com", ep),
+ "https://api.openai.com/v1/embeddings"
);
assert_eq!(
- normalize_openai_base("https://api.openai.com/"),
- "https://api.openai.com"
+ normalize_openai_base("https://openrouter.ai/api/v1", ep),
+ "https://openrouter.ai/api/v1/embeddings"
);
- // OpenRouter's docs / ai-memory's OpenRouter env entry.
assert_eq!(
- normalize_openai_base("https://openrouter.ai/api/v1"),
- "https://openrouter.ai/api"
+ normalize_openai_base("http://localhost:11434/v1", ep),
+ "http://localhost:11434/v1/embeddings"
);
- // Ollama's openai-compat path (`docker logs` shows it
- // canonically advertised this way).
assert_eq!(
- normalize_openai_base("http://localhost:11434/v1"),
- "http://localhost:11434"
+ normalize_openai_base("https://example.com/v123", ep),
+ "https://example.com/v123/v1/embeddings"
);
- // Don't accidentally strip a `/v1` that's part of a longer
- // segment (e.g. `/v123/`).
assert_eq!(
- normalize_openai_base("https://example.com/v123"),
- "https://example.com/v123"
+ normalize_openai_base("https://api.z.ai/api/coding/paas/v4", ep),
+ "https://api.z.ai/api/coding/paas/v4/embeddings"
);
}
}