DAOCALM developer integration examples
Documentation: https://www.daocalm.com/docs
Updated: 2026-10-24

These helpers call actual provider/runtime APIs on infrastructure you configure.
Enterprise Provisioned Access is arranged for an agreed deployment scope.
The helpers use the AWS Python SDK and a Python client for your ComfyUI runtime;
they do not define a DAOCALM-hosted REST endpoint or packaged SDK. No API key is
embedded. The examples do not grant an account, service entitlement or SLA.

REQUIREMENTS
Python 3.10 or later. In a virtual environment:
  python -m pip install boto3 requests
The libraries above are the AWS Python SDK and the Requests HTTP client.
Pin versions tested with your own deployment before production use.

NEXUS / AWS BEDROCK
1. Configure AWS credentials through a supported SDK credential provider, such
   as an authorized role or a locally configured AWS profile. Do not put keys
   in source code or website forms.
2. Choose a region and a model or inference profile that your AWS account can
   use with Converse. The role requires bedrock:InvokeModel on its resources.
3. Save a business brief as brief.txt (at most 12,000 characters).
4. Run, replacing the values with your actual configuration:
   python nexus_bedrock.py --region YOUR_REGION --model YOUR_MODEL_ID --input brief.txt --output nexus-result.json
5. Review nexus-result.json. The helper performs two specialist reviews and one
   synthesis. It never executes generated instructions. AWS charges may apply.
   --workers 1 reduces concurrent calls. The default is two, not eight.

The SDK signs requests. Per-attempt timeouts are 5 seconds to connect and 120
seconds to read; standard retry mode allows at most 3 total attempts per call.
A whole job has no provider idempotency guarantee and can cost more if rerun.
No ChromaDB collection, external tool access, memory store or proprietary Nexus
server is created by this example.

CINESYNTH / COMFYUI
1. Run a trusted, authorized ComfyUI installation with the needed models and
   nodes. Validate a small workflow through its UI before calling the API.
2. Export that working workflow through File > Export (API), saving it as
   workflow_api.json. A visual editor graph is a different format. No generic
   workflow bundled here can guarantee video output or 4K compatibility.
3. Use local loopback http://127.0.0.1:8188, or an HTTPS URL protected by the
   authentication and network controls your operator has configured.
4. If your HTTPS gateway explicitly supports Bearer tokens, set
   COMFY_GATEWAY_BEARER_TOKEN securely in your process environment. This is
   gateway authentication, not a built-in ComfyUI or DAOCALM key. If your
   gateway uses other headers or client certificates, adapt the session first.
5. Run against your local runtime:
   python cinesynth_comfyui.py --workflow workflow_api.json --output-dir cinesynth-output
   For a remote runtime add --base-url https://YOUR_RUNTIME_HOST
6. The helper submits once, records prompt_id, polls history, then downloads
   compatible file descriptors through /view. Default polling timeout: 900
   seconds. Per-artifact cap: 512 MiB. Maximum artifact count: 20.
7. If polling times out, the server job may continue. Check job.json and resume
   without submitting the workflow again, using a NEW output directory:
   python cinesynth_comfyui.py --prompt-id YOUR_PROMPT_ID --output-dir resumed-output
   Supply the same --base-url and authentication as the original request.

Submission is not retried automatically. If the POST response is lost, inspect
the runtime queue/history before submitting again. Read-only polling retries
connection errors and HTTP 429/502/503/504 until its deadline. A successful
history record must include status.completed=true and status.status_str=success.
Empty history alone does not distinguish queued, running, expired or unknown.

The downloader recognizes filename/subfolder/type metadata in output arrays.
Video nodes with other response formats need their own adapter. The helper
does not convert an image workflow into a video workflow. Workflow-specific
duration, resolution, frame rate, models, seeds and licenses must be validated.

Data and permissions
Run only trusted workflows and custom nodes. They can use local resources or
external services. Obtain permission for source media, likenesses, brands,
models and music. Review every output before publication. Keep access tokens
out of exported workflow files. Store outputs/history in an access-controlled
location; history can include workflow inputs. Control provider spend and
retention in your own deployment.

Validation boundary
These downloads are client examples. They do not supply a DAOCALM backend,
production credentials, models or GPU service. Provider documentation and
offline client behavior were checked; a live AWS inference or ComfyUI GPU
render was not run for this website release.

Primary references
https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_Converse.html
https://docs.aws.amazon.com/boto3/latest/reference/services/bedrock-runtime/client/converse.html
https://docs.aws.amazon.com/boto3/latest/guide/credentials.html
https://docs.aws.amazon.com/boto3/latest/guide/retries.html
https://docs.comfy.org/development/comfyui-server/comms_routes
https://github.com/Comfy-Org/ComfyUI/blob/master/script_examples/basic_api_example.py
https://github.com/Comfy-Org/ComfyUI/blob/master/server.py