From d11c2d25897e0036955d4e43d40b7f594531158c Mon Sep 17 00:00:00 2001
From: Jacob <jacob@katniss.top>
Date: Mon, 19 Oct 2026 15:56:06 -0400
Subject: [PATCH] feat: cohere + more models

---
 src/lib.rs                                  |  3 ++
 src/server/chat/completions.rs              | 20 +++++++++
 src/server/embeddings.rs                    | 50 +++++++++++++++++++++
 src/server/frontend/index.html              | 12 +++++
 src/server/{frontend.rs => frontend/mod.rs} |  0
 5 files changed, 85 insertions(+)
 create mode 100644 src/server/frontend/index.html
 rename src/server/{frontend.rs => frontend/mod.rs} (100%)

diff --git a/src/lib.rs b/src/lib.rs
index 87a8745..5b8f0ff 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -39,6 +39,9 @@ pub struct AppArgs {
     #[arg(long)]
     pub cloudflare_workers_ai_credentials: Vec<CloudflareCredentials>,
 
+    #[arg(long)]
+    pub cohere_api_key: Vec<SecretString>,
+
     #[arg(long)]
     pub google_ai_studio_api_key: Vec<SecretString>,
 
diff --git a/src/server/chat/completions.rs b/src/server/chat/completions.rs
index e94f8c0..5300e70 100644
--- a/src/server/chat/completions.rs
+++ b/src/server/chat/completions.rs
@@ -88,6 +88,7 @@ impl Display for ShortcutAndMaybeReasoning {
 #[derive(Debug, Clone)]
 enum Provider {
     CloudflareWorkersAi,
+    Cohere,
     GoogleAiStudio,
     Groq,
     HuggingFace,
@@ -123,6 +124,7 @@ impl Shortcut {
                 (P::OpenRouter, "poolside/laguna-xs-2.1:free".into()),
                 (P::ZAi, "glm-4.7-flash".into()),
                 (P::NvidiaNim, "google/diffusiongemma-26b-a4b-it".into()),
+                (P::Cohere, "north-mini-code-1-0".into()),
                 (P::OpenRouter, "cohere/north-mini-code:free".into()),
                 (P::NvidiaNim, "nvidia/nemotron-3.5-lightning-30b-a3b".into()),
                 (
@@ -207,6 +209,7 @@ impl Shortcut {
                 ),
                 (P::VercelAiGateway, "mistral/mistral-large-3".into()),
                 (P::Ollama, "nemotron-3-super".into()),
+                (P::Cohere, "command-a-plus-05-2026".into()),
                 (P::NvidiaNim, "nvidia/nemotron-3-super-120b-a12b".into()),
                 (P::Kilo, "nvidia-nemotron-3-super-120b-a12b-free".into()),
                 (
@@ -233,6 +236,7 @@ impl Shortcut {
                 (P::NvidiaNim, "deepseek-ai/deepseek-v4.1-flash".into()),
                 (P::HuggingFace, "deepseek-ai/DeepSeek-V4.1-Flash".into()),
                 (P::OpenCode, "deepseek-v4.1-flash-free".into()),
+                (P::OpenRouter, "minimax/minimax-m3.1-flash:free".into()),
                 (P::VercelAiGateway, "inclusionai/ling-3.1-flash".into()),
                 (P::HuggingFace, "inclusionAI/Ling-3.1-flash".into()),
                 (P::Ollama, "beam".into()),
@@ -395,6 +399,22 @@ pub async fn create_chat_completion(
                             .build()
                     })
                     .collect(),
+                Provider::Cohere => state
+                    .app_args
+                    .cohere_api_key
+                    .iter()
+                    .cloned()
+                    .map(|api_key| {
+                        Upstream::builder()
+                            .url(
+                                "https://api.cohere.com/compatibility/v1/chat/completions"
+                                    .parse()
+                                    .expect("TODO"),
+                            )
+                            .api_key(api_key)
+                            .build()
+                    })
+                    .collect(),
                 Provider::Groq => state
                     .app_args
                     .groq_api_key
diff --git a/src/server/embeddings.rs b/src/server/embeddings.rs
index a8ab903..53a0056 100644
--- a/src/server/embeddings.rs
+++ b/src/server/embeddings.rs
@@ -34,6 +34,18 @@ pub enum Shortcut {
     #[strum(serialize = "codestral-embed-2505")]
     CodestralEmbed2505,
 
+    #[strum(serialize = "cohere-embed-v4.0")]
+    CohereEmbedV4_0,
+
+    #[strum(serialize = "cohere-embed-v5.0-fast")]
+    CohereEmbedV5_0Fast,
+
+    #[strum(serialize = "cohere-embed-v5.0-pro")]
+    CohereEmbedV5_0Pro,
+
+    #[strum(serialize = "gemini-embedding-2")]
+    GeminiEmbedding2,
+
     #[strum(serialize = "lfm2.5-embedding-350m")]
     Lfm2_5Embedding350M,
 
@@ -46,6 +58,8 @@ pub enum Shortcut {
 
 #[derive(Debug, Clone)]
 enum Provider {
+    Cohere,
+    GoogleAiStudio,
     Mistral,
     NvidiaNim,
     OpenRouter,
@@ -58,6 +72,10 @@ impl Shortcut {
 
         match self {
             S::CodestralEmbed2505 => vec![(P::Mistral, "codestral-embed-2505".into())],
+            S::CohereEmbedV4_0 => vec![(P::Cohere, "embed-v4.0".into())],
+            S::CohereEmbedV5_0Fast => vec![(P::Cohere, "embed-v5.0-fast".into())],
+            S::CohereEmbedV5_0Pro => vec![(P::Cohere, "embed-v5.0-pro".into())],
+            S::GeminiEmbedding2 => vec![(P::GoogleAiStudio, "gemini-embedding-2".into())],
             S::Lfm2_5Embedding350M => {
                 vec![(P::OpenRouter, "liquid/lfm-2.5-embedding-350m:free".into())]
             }
@@ -103,6 +121,22 @@ pub async fn create_embeddings(
 
     for (provider, model) in shortcut.suggestions() {
         let upstreams: Vec<Upstream> = match provider {
+            Provider::Cohere => state
+                .app_args
+                .cohere_api_key
+                .iter()
+                .cloned()
+                .map(|api_key| {
+                    Upstream::builder()
+                        .url(
+                            "https://api.cohere.com/compatibility/v1/embeddings"
+                                .parse()
+                                .expect("TODO"),
+                        )
+                        .api_key(api_key)
+                        .build()
+                })
+                .collect(),
             Provider::Mistral => state
                 .app_args
                 .mistral_api_key
@@ -119,6 +153,22 @@ pub async fn create_embeddings(
                         .build()
                 })
                 .collect(),
+            Provider::GoogleAiStudio => state
+                .app_args
+                .google_ai_studio_api_key
+                .iter()
+                .cloned()
+                .map(|api_key| {
+                    Upstream::builder()
+                        .url(
+                            "https://generativelanguage.googleapis.com/v1beta/openai/embeddings"
+                                .parse()
+                                .expect("TODO"),
+                        )
+                        .api_key(api_key)
+                        .build()
+                })
+                .collect(),
             Provider::NvidiaNim => state
                 .app_args
                 .nvidia_nim_api_key
diff --git a/src/server/frontend/index.html b/src/server/frontend/index.html
new file mode 100644
index 0000000..00543ba
--- /dev/null
+++ b/src/server/frontend/index.html
@@ -0,0 +1,12 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>Basic Page</title>
+  </head>
+  <body>
+    <h1>Hello, world!</h1>
+    <p>This is a basic HTML page.</p>
+  </body>
+</html>
diff --git a/src/server/frontend.rs b/src/server/frontend/mod.rs
similarity index 100%
rename from src/server/frontend.rs
rename to src/server/frontend/mod.rs