From 9c480aaac2ccc0e19ed1344d1172d35168062e93 Mon Sep 17 00:00:00 2001
From: James Barker <james@jcbdevelopment.co.uk>
Date: Mon, 10 Feb 2025 09:13:55 +0000
Subject: [PATCH 1/2] add retries to openrouter generation endpoint to prevent
 failures on slow usage stat updates server side

Signed-off-by: James Barker <james@jcbdevelopment.co.uk>
---
 src/api/providers/openrouter.ts | 51 +++++++++++++++++----------------
 1 file changed, 27 insertions(+), 24 deletions(-)

diff --git a/src/api/providers/openrouter.ts b/src/api/providers/openrouter.ts
index 0e23c5d35d1..633b6baa52d 100644
--- a/src/api/providers/openrouter.ts
+++ b/src/api/providers/openrouter.ts
@@ -183,31 +183,34 @@ export class OpenRouterHandler implements ApiHandler, SingleCompletionHandler {
 			// }
 		}
 
-		await delay(500) // FIXME: necessary delay to ensure generation endpoint is ready
-
-		try {
-			const response = await axios.get(`https://openrouter.ai/api/v1/generation?id=${genId}`, {
-				headers: {
-					Authorization: `Bearer ${this.options.openRouterApiKey}`,
-				},
-				timeout: 5_000, // this request hangs sometimes
-			})
+		// retry fetching generation details
+		let retries = 0
+		while (retries < 3) {
+			await delay(500) // FIXME: necessary delay to ensure generation endpoint is ready
+			try {
+				const response = await axios.get(`https://openrouter.ai/api/v1/generation?id=${genId}`, {
+					headers: {
+						Authorization: `Bearer ${this.options.openRouterApiKey}`,
+					},
+					timeout: 5_000, // this request hangs sometimes
+				})
 
-			const generation = response.data?.data
-			console.log("OpenRouter generation details:", response.data)
-			yield {
-				type: "usage",
-				// cacheWriteTokens: 0,
-				// cacheReadTokens: 0,
-				// openrouter generation endpoint fails often
-				inputTokens: generation?.native_tokens_prompt || 0,
-				outputTokens: generation?.native_tokens_completion || 0,
-				totalCost: generation?.total_cost || 0,
-				fullResponseText,
-			} as OpenRouterApiStreamUsageChunk
-		} catch (error) {
-			// ignore if fails
-			console.error("Error fetching OpenRouter generation details:", error)
+				const generation = response.data?.data
+				console.log("OpenRouter generation details:", response.data)
+				yield {
+					type: "usage",
+					// cacheWriteTokens: 0,
+					// cacheReadTokens: 0,
+					// openrouter generation endpoint fails often
+					inputTokens: generation?.native_tokens_prompt || 0,
+					outputTokens: generation?.native_tokens_completion || 0,
+					totalCost: generation?.total_cost || 0,
+					fullResponseText,
+				} as OpenRouterApiStreamUsageChunk
+			} catch (error) {
+				// ignore if fails
+				console.error("Error fetching OpenRouter generation details:", error)
+			}
 		}
 	}
 	getModel(): { id: string; info: ModelInfo } {

From b85960ba2ae6d33ca93598e9ba051d17919ada1e Mon Sep 17 00:00:00 2001
From: James Barker <james@jcbdevelopment.co.uk>
Date: Mon, 10 Feb 2025 14:58:44 +0000
Subject: [PATCH 2/2] updated to fix no break condition on loop and to use the
 same number of retries and delay as the glama provider (for consistency)

Signed-off-by: James Barker <james@jcbdevelopment.co.uk>
---
 src/api/providers/openrouter.ts | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/api/providers/openrouter.ts b/src/api/providers/openrouter.ts
index 633b6baa52d..5d6b79577fa 100644
--- a/src/api/providers/openrouter.ts
+++ b/src/api/providers/openrouter.ts
@@ -184,9 +184,9 @@ export class OpenRouterHandler implements ApiHandler, SingleCompletionHandler {
 		}
 
 		// retry fetching generation details
-		let retries = 0
-		while (retries < 3) {
-			await delay(500) // FIXME: necessary delay to ensure generation endpoint is ready
+		let attempt = 0
+		while (attempt++ < 10) {
+			await delay(200) // FIXME: necessary delay to ensure generation endpoint is ready
 			try {
 				const response = await axios.get(`https://openrouter.ai/api/v1/generation?id=${genId}`, {
 					headers: {
@@ -207,6 +207,7 @@ export class OpenRouterHandler implements ApiHandler, SingleCompletionHandler {
 					totalCost: generation?.total_cost || 0,
 					fullResponseText,
 				} as OpenRouterApiStreamUsageChunk
+				return
 			} catch (error) {
 				// ignore if fails
 				console.error("Error fetching OpenRouter generation details:", error)