Deploy Flowise

Flowise is a visual workflow builder for language-model applications. Moltern deploys it with a managed PostgreSQL dependency, persistent workspace storage, a protected administrator account and a live HTTPS URL.

What Moltern deploys​

Component	Purpose	Access
Flowise 3.1.4	Visual chatflow editor and prediction API	Public HTTPS URL with Flowise sign-in
Managed PostgreSQL	Accounts, variables, chatflows and execution metadata	Private to the environment
Workspace files	Flowise files that must survive runtime replacement	Assigned service path

The database is created and connected automatically. It does not receive a public URL or a dedicated disk claim.

Deploy Flowise​

1. Open Services and choose Flowise.
2. Select the project and environment.
3. Enter the administrator name and email that should own the first account.
4. Set a strong administrator password.
5. Review the install impact, then choose Deploy service.

The install preview should show two workloads: Flowise and its managed database. It should not request a dedicated volume because both workloads use their assigned paths in the workspace filespace.

The first start runs database migrations and can take longer than a normal restart. Wait until the service status is Running before opening its URL.

Sign in​

Open the generated URL and sign in with the administrator email and password entered during deployment. There is no public first-owner registration window.

You can retrieve the masked setup values later from Service → Settings → Connection details. Moltern requires account confirmation before revealing sensitive values.

Connect private Ollama​

Deploy Ollama in the same environment before continuing. In Flowise:

1. Open Settings in the Moltern service page.
2. Under Private service connections, select the Ollama service.
3. Choose Connect.
4. Wait while Moltern replaces the Flowise runtime and returns it to Running.

Moltern grants only the selected connection, creates scoped variables and adds the corresponding private network policy. Ollama remains without a public URL.

Use the connection in Flowise​

The connection card displays the generated variable names. To use the URL without copying a private address into a flow:

1. In Flowise, open Settings → Variables.
2. Create a variable with the generated name ending in _URL.
3. Set its type to Runtime and leave its stored value empty.
4. Add a ChatOllama node to a chatflow.
5. Set Base URL to {{$vars.YOUR_OLLAMA_URL_VARIABLE}}.
6. Set Model Name to a model already installed in Ollama.
7. Connect ChatOllama to the chain and save the flow.

The runtime variable reads the scoped value supplied by Moltern. It does not store the private endpoint in the Flowise database.

The production E2E used qwen2.5:1.5b and verified a real prediction through the private attachment. A healthy page or pod alone was not treated as proof.

Persistence and operations​

Use Stop service and Start service on the Moltern service page for a controlled runtime replacement. The administrator account, runtime variable, chatflow and post-restart Ollama prediction were all retained in the production test.

Changing a private connection also replaces the Flowise runtime. Plan a short interruption because Flowise is currently deployed as one replica and its final upstream image is approximately 1.5 GB. A cold pull to a worker can take several minutes; subsequent starts on a cached worker are faster.

Storage and metering​

Moltern measures Flowise and its managed database separately even though both use the workspace filespace. In the production validation:

• Flowise reported 31,848 bytes.
• Managed PostgreSQL reported 49,831,956 bytes.
• Resource accounting reported 100 mCPU, 512 MiB and one Flowise replica.
• PostgreSQL accounting reported 100 mCPU, 256 MiB and one replica.

The collector values matched direct measurements at the same collection point. Elapsed invoice and proration certification remains tracked separately from this point-in-time resource test.

Delete Flowise​

Choose Delete Service, select the option that removes stored data and confirm the protected action. Moltern removes:

• the Flowise runtime and route;
• the managed PostgreSQL dependency;
• both assigned workspace paths;
• generated secrets and private-connection policy; and
• active CPU, memory, replica and storage usage.

The workspace filespace itself remains available for the team's other workloads.

Known limitations​

The final upstream image starts and ChatOllama works, but its startup log reports upstream dependency failures for the ReAct Agent Chat, ReAct Agent LLM and AWS Bedrock nodes. Do not depend on those nodes in this catalog version. The Moltern validation did not cover external model-provider credentials, high availability, load performance, backup restore or interrupted migration recovery.

Troubleshooting​

Symptom	What to check
First start takes several minutes	Check Live Logs for image pull and database migration progress.
Sign-in fails	Confirm the administrator email and reveal the managed setup values through the protected connection-details flow.
ChatOllama cannot connect	Confirm Ollama is Running in the same environment, the private connection is present and the Flowise variable type is Runtime.
Model not found	Pull the exact model in Ollama and use the same model name in ChatOllama.
A node is absent from the editor	Review Known limitations; several nodes are broken in the archived final upstream release.
Restart loses a flow	Confirm the managed database is Running and contact support before deleting or redeploying stored data.

Related guides​

• Ollama
• AnythingLLM
• All services