From 717f4c30ed5853deac60b2d68e7013701c5fb8ac Mon Sep 17 00:00:00 2001 From: autoimago Date: Thu, 8 Oct 2026 11:50:04 +0200 Subject: [PATCH 1/2] =?UTF-8?q?=F0=9F=93=9D=20docs:=20Add=20Gonka=20DAHL?= =?UTF-8?q?=20custom=20endpoint?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../ai_endpoints/gonka-dahl.mdx | 44 +++++++++++++++++++ 1 file changed, 44 insertions(+) create mode 100644 content/docs/configuration/librechat_yaml/ai_endpoints/gonka-dahl.mdx diff --git a/content/docs/configuration/librechat_yaml/ai_endpoints/gonka-dahl.mdx b/content/docs/configuration/librechat_yaml/ai_endpoints/gonka-dahl.mdx new file mode 100644 index 000000000..181cd6566 --- /dev/null +++ b/content/docs/configuration/librechat_yaml/ai_endpoints/gonka-dahl.mdx @@ -0,0 +1,44 @@ +--- +title: Gonka DAHL +icon: Network +description: Configure Gonka DAHL as a custom endpoint in LibreChat. +--- + +Gonka DAHL is an OpenAI-compatible inference API for open-weight models served on the Gonka decentralized GPU network. A new account gets 100 million free tokens, and sign-up asks for a username only. + +## Get an API key + +1. Create an account at [inference.dahl.global](https://inference.dahl.global) and save the fingerprint it shows. It works as your password and cannot be recovered. +2. On the [account page](https://inference.dahl.global/account), allocate tokens from your pool to an API key. New keys start with 0 tokens. +3. Add the key to your `.env` file: + +```bash filename=".env" +DAHL_API_KEY=your-api-key +``` + +## Configuration + +Add the endpoint under `endpoints.custom` in your `librechat.yaml`: + +```yaml filename="librechat.yaml" + - name: "Gonka DAHL" + apiKey: "${DAHL_API_KEY}" + baseURL: "https://inference.dahl.global/v1" + models: + default: [ + "MiniMaxAI/MiniMax-M2.7", + "deepseek-ai/DeepSeek-V4-Flash-0731", + "zai-org/GLM-5.3-Flash" + ] + fetch: true + titleConvo: true + titleModel: "current_model" + modelDisplayLabel: "Gonka DAHL" +``` + +## Notes + +- `fetch: true` loads the live list from `GET /v1/models`, and the `default` list is the fallback. Model IDs change over time, so check the [models docs](https://inference.dahl.global/docs/models/). +- MiniMax M2.7 returns its reasoning inside `` tags in the message content. +- Busy models can return a `model_concurrency` error at peak times, and signed-in and paid accounts are admitted first. Switch to another model or retry. +- Top-ups are paid from the account page. See [Tokens & keys](https://inference.dahl.global/docs/tokens/). From b040ecec7a2c46c1d0071e006796b8ad6cd6bca6 Mon Sep 17 00:00:00 2001 From: autoimago Date: Thu, 8 Oct 2026 11:50:37 +0200 Subject: [PATCH 2/2] =?UTF-8?q?=F0=9F=93=9D=20docs:=20List=20Gonka=20DAHL?= =?UTF-8?q?=20in=20AI=20endpoints=20navigation?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- content/docs/configuration/librechat_yaml/ai_endpoints/meta.json | 1 + 1 file changed, 1 insertion(+) diff --git a/content/docs/configuration/librechat_yaml/ai_endpoints/meta.json b/content/docs/configuration/librechat_yaml/ai_endpoints/meta.json index c46fe3dfd..a2ee7a6e7 100644 --- a/content/docs/configuration/librechat_yaml/ai_endpoints/meta.json +++ b/content/docs/configuration/librechat_yaml/ai_endpoints/meta.json @@ -10,6 +10,7 @@ "databricks", "deepseek", "fireworks", + "gonka-dahl", "groq", "helicone", "huggingface",