Add detection
Scope OpenDetector AI-text review or prompt protection, with a separate detection contract and an operating policy.
Plan detection accessAgent API & Grid / Developers
Build your first agent.
Start with one bounded task.
Request Agent API access, confirm an enabled agent and run a synthetic task with an explicit budget. Grid model API examples are available below.
Agent API and Grid access are by invitation. Your account determines enabled services and effective rates.
Agent API / First run
Delegate, inspect, then review.
Use the API origin, bearer key and enabled agent ID confirmed during onboarding. Confirm its tools and approval policy before sending a request. The example uses synthetic input and an illustrative budget.
- Set
FORWARD_AGENT_BASEto the supplied origin, without a trailing slash. KeepFORWARD_AGENT_KEYin your server or local environment. - Create
first-agent.jsonwith your enabled agent ID, a goal and budget as shown in the guide. Run from a private directory withumask 077. - A successful delegation returns HTTP 202. Read
run_idandstatusfromrun.json. Keep returned tokens private. - Inspect
GET /v1/agent/runs/{run_id}with the same authorization. Review completion, output and usage before increasing the scope.
Delegate an agent run
curl --fail-with-body --silent --show-error \
-X POST "${FORWARD_AGENT_BASE}/v1/agent/delegate" \
-H "Authorization: Bearer ${FORWARD_AGENT_KEY}" \
-H "Content-Type: application/json" \
--data-binary @first-agent.json \
--output run.jsonTemplate only. Your account controls availability and effective rates. Sending a request can consume usage; a budget is not an estimate of the result's cost.
03 / First call
Choose your client.
Keep the same endpoint.
The examples read the key and model ID from your environment. They send one chat message and print the reply.
Before you run
Use an invited account, FORLEX_API_KEY and a chat-capable GRID_MODEL_ID from your enabled model list.
- API base
- https://api.genai.forlex.cloud/v1
- Request
- POST /chat/completions
A successful request confirms access. Your evaluation determines whether a model fits the workflow.
Plan the evaluationInstall the SDK: python3 -m pip install openai. Save the example as first-call.py, then run python3 first-call.py.
first-call.py
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.genai.forlex.cloud/v1",
api_key=os.environ["FORLEX_API_KEY"],
)
response = client.chat.completions.create(
model=os.environ["GRID_MODEL_ID"],
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)Access and setup
Set up your account and model.
Grid is by invitation. Keep your key in your local environment or server configuration, outside browser code and source control. The examples use placeholders.
01 / Access and key
Once invited, open the Console, go to Keys and create an API key.
Set your API key
Set your API key
export FORLEX_API_KEY="replace-with-your-key"02 / Enabled model
Set FORLEX_API_KEY first. Choose a model that supports chat completions from your account list and use its exact ID.
List and select an enabled model
Replace both placeholders before running the first call.
List enabled models
curl https://api.genai.forlex.cloud/v1/models \
-H "Authorization: Bearer ${FORLEX_API_KEY:?Set FORLEX_API_KEY first}"Choose an enabled model
export GRID_MODEL_ID="replace-with-an-enabled-chat-model-id"04 / Prepare for production
Understand the response. Review the cost.
Add error handling, measure useful results and check consumption in the Console. Confirm data, deployment and operating requirements with the people who approve production.
Response troubleshooting
- 401 / 403
- Check the key, invitation and account access. An enabled model list is specific to your account.
- Access and keys
- 402 / 429
- Review billing or request limits. Use backoff for rate limits; resolve billing before retrying.
- Billing and limits
- 503
- An idle serverless lane can be warming up. Follow Retry-After when present and use bounded retries.
- Errors and cold start
Take the next step with your team.
Connect your coding tool
Prepare a tool-specific configuration, then evaluate a pinned model on bounded coding tasks with your team.
Prepare a Connect setupAdapt a model
Define the task, data readiness and evaluation before choosing a checkpoint or training method.
Explore trainingDeliver a workflow
Bring the process and integrations. We scope the evaluation and implementation with your team.
Explore Projects