Commit Graph

25 Commits

Author SHA1 Message Date
00a7fba22b infra: Expose Ollama service via Traefik
Adds the `traefik.enable=true` tag to the `ollama` service definition within the `ai-backend` Nomad job.

This enables Traefik to automatically discover and route incoming requests to the Ollama service, making it accessible through the Traefik ingress.
2025-12-29 00:46:46 -05:00
7eab52e5bc fix(ai-backend): Specify OLLAMA_HOST with explicit port
The `OLLAMA_HOST` environment variable in the `ai-backend` Nomad job has been updated from `0.0.0.0` to `0.0.0.0:11434`.

This change ensures OLLAMA explicitly binds to the correct address and port, preventing potential connectivity issues or misconfigurations where the port might be omitted or incorrectly inferred.
2025-12-29 00:19:26 -05:00
437c81b9e6 chore(ai): Remove LobeChat and simplify Ollama URL
Removed the `lobechat` group and its associated configuration from `stacks/ai/ai-frontend.nomad` as this application is no longer required.

Updated the `OLLAMA_BASE_URL` for the `ai-frontend` (OpenWebUI) task to use the concise `ollama` service name (`http://ollama:11434`). This leverages Nomad's built-in service discovery more efficiently, simplifying the URL from `http://ollama.service.consul:11434`.

Additionally, the `memory` allocation for the `ai-frontend` task was increased from `1024` to `4000` to provide more stable resources for the OpenWebUI application.
2025-12-28 23:48:04 -05:00
518aaab5f1 ci(deploy): Add 'newt' deploy option and make stack selection explicit
The `newt` stack has been added to the list of available deployment options for the Gitea Actions workflow.
Additionally, the default value for the `stack` input has been removed, making it a mandatory selection. This ensures that users explicitly choose which service to deploy, preventing unintended deployments and improving clarity.
2025-12-28 23:09:24 -05:00
dc28f64154 feat(networking): Deploy newt agent as a Nomad service
Introduces a new Nomad job definition (`newt.nomad`) to deploy the `newt` agent as a managed service within the Homelab environment.

This job configures a single `newt-agent` task using the `fosrl/newt:latest` Podman image. The agent is set up with specific environment variables for endpoint, ID, and secret, enabling it to connect to the Pangolin service and establish network connectivity. Resource limits for CPU and memory are also defined.
2025-12-28 23:08:38 -05:00
93876ad444 refactor(nomad): Centralize DNS configuration for AI frontends
Moves DNS server and search domain configuration from task `config` blocks to the job/group `network` blocks for `ai-frontend` (OpenWebUI) and `lobechat`.

This refactoring centralizes DNS settings for all tasks within a Nomad job or group, ensuring consistent DNS resolution and aligning with recommended Nomad configuration patterns.
2025-12-28 10:57:15 -05:00
00de9d60fa fix(ai): Adjust LobeChat health check and refine Nomad configurations
Corrected the LobeChat service health check path from `/api/health` to `/` as the root endpoint is the correct health indicator. This resolves an issue where the service was not being properly marked as healthy by Nomad.

Additionally, this commit refactors comments across both `ai-frontend` (OpenWebUI) and `lobechat` jobs within the `ai-frontend.nomad` file. Redundant or outdated comments regarding DNS "reverts," version pinning, and security notes have been removed or simplified to improve readability and reflect the stable state of the configurations. The Consul DNS configuration is now consistently marked as a "DNS Fix" rather than a temporary solution.
2025-12-28 00:14:12 -05:00
2958fece21 fix(ai-frontend): Configure Consul DNS for service discovery
This commit addresses service discovery issues by explicitly configuring Consul DNS for the `openwebui` and `lobechat` tasks within the `ai-frontend` job.

Previously, services were unable to reliably resolve `ollama.service.consul`. This is resolved by:

*   Setting `dns_servers` to the Consul server IP (`192.168.1.133`).
*   Setting `dns_search` to `service.consul`.
*   Reverting `OLLAMA_BASE_URL` and `OLLAMA_PROXY_URL` environment variables to use `ollama.service.consul` as intended.
*   The `lobechat` task's image is now pinned to `v1.143.0` for improved stability.
*   Removed outdated comments regarding host pinning.
2025-12-28 00:11:55 -05:00
d57cee3e48 infra(ai-frontend): Update Nomad job hostname constraint to hp1-home
Updates the hostname constraint in the `ai-frontend` Nomad job from `hp-mini-1` to `hp1-home` for both `openwebui` and `lobechat` tasks. This correction ensures the services are pinned to the accurate target host.

Removed outdated `FIXED:` comments for cleanup.
2025-12-27 23:40:12 -05:00
1957f15586 refactor(ai-frontend): Improve Nomad job configuration and clarity
*   Reformatted the `resources` block for the `openwebui` group to a multi-line format to resolve an illegal semicolon syntax issue.
*   Added explicit comments to both `openwebui` and `lobechat` groups to clarify the host pinning constraint.
*   Streamlined comments in the `lobechat` group's `resources` block to concisely state the memory allocation, removing outdated "fix" comments and adding a current "FIXED" comment.

These changes enhance the readability, maintainability, and correctness of the Nomad job definition.
2025-12-27 23:36:46 -05:00
aab5e07310 fix(ai-frontend): Prevent LobeChat OOM and update host constraints
This commit addresses several issues and updates in the `ai-frontend.nomad` job configuration:

- **LobeChat OOM Fix**: Increased `lobechat` service memory allocation from 512MB to 1024MB to prevent out-of-memory crashes and improve stability.
- **Node Constraint Update**: Updated the `attr.unique.hostname` constraint for both `openwebui` and `lobechat` groups from `hp1-home` to `hp-mini-1` to reflect the correct node hostname.
- **Static Port Assignment**: Explicitly added `static` port definitions for `http` in both `openwebui` (8080) and `lobechat` (3210) services to ensure consistent port binding.
- **Minor Housekeeping**: Cleaned up comments and formatted resource blocks for better readability.
2025-12-27 23:33:08 -05:00
c0f67aef53 Change hostnames 2025-12-27 15:12:07 -05:00
9d043bd142 Frontend 2025-12-27 15:09:05 -05:00
c69943692b sdfsf 2025-12-27 14:51:58 -05:00
762bfb29fa adfadfs 2025-12-27 14:49:00 -05:00
619124a30c fgsfg 2025-12-27 14:45:58 -05:00
99b50a5902 Fix brackets 2025-12-27 14:42:03 -05:00
a6774d816d Maybe Fixed GPU? 2025-12-27 14:37:33 -05:00
b104dc4640 Fixed eGPU deploy? 2025-12-27 14:26:56 -05:00
2215d8c2e8 Corrected directory structure 2025-12-27 13:40:07 -05:00
a3df4cdadb Fixed setup script 2025-12-27 13:33:17 -05:00
aebff3da61 rdtyu 2025-12-27 12:51:00 -05:00
57ef5a13b1 Add AI stack 2025-12-27 12:43:54 -05:00
2b5d5d9df2 Changed Deploy.yml 2025-12-27 12:35:58 -05:00
6d73d9365c Initial commit for Gitea Workflows 2025-12-27 12:27:56 -05:00