Setting Up an IBM i Environment
How to set up a CoderFlow environment for an IBM i application served by Profound UI Genie. Unlike the web-application environments, nothing runs in the task container: CoderFlow proxies to the running Genie server, so a task can drive real 5250 green-screen and Rich Display Web UI sessions while the agent edits and compiles source on the IBM i.
For the full connection reference and the IBM i-side setup, see IBM i connections; for the bigger picture, see the IBM i overview.
How this differs from a web-app environment
- It isn't a web app. The UI is a green-screen (5250) or Rich Display IBM i application rendered by a running Genie server — not a backend serving an SPA.
- No app runtime to install and no Start Command. The container holds your source for editing and compiling; the live UI comes from Genie, not from a process in the container.
- The proxy is configured by a connection, not by hand. An IBM i Interactive Sessions connection points the Application Server proxy at Genie automatically.
- Compilation happens on the IBM i system. The agent compiles with
codermakeinto an isolated per-task build library, rather than building in the container.
What you'll need
- A reachable IBM i system running Profound UI / Genie Version 6 Fix Pack 37.0 or later.
- HTTPS on the selected Genie HTTP instance, with a trusted certificate matching the connection hostname, for captured Genie screen renderings. Configure this through IBM HTTP Server administration / Digital Certificate Manager.
- Both CoderFlow
SetEnvdirectives enabled globally in that instance’s/www/<instance>/conf/httpd.conf(shown below). - Profound Logic RAS installed (used by the
sqlskill) and an IBM i user profile configured with the right library list and<product-library>/PUISETENVcalled as or from its initial program. Use the selected installation’s actual product library; adding it to LIBL alone is insufficient. Custom initial programs may call PUISETENV indirectly. The connection form's How to set up IBM i user profile helper lists the exact requirements — see IBM i connections. - Permission to create environments and connections in the CoderFlow Web UI.
Create the environment
1. Basic settings
Go to Environments → New Environment. On the Overview tab set a Description, a Default Agent, a Docker Image name, and a Timezone. You do not need any Pre-clone instructions — the IBM i build tooling (codermake) is already in the base image, and the application UI is proxied.
2. Add the source repository
On the Repositories tab, add the repository holding your RPG/DDS/Profound UI source. To bring existing members in and generate compile rules, enable IBM i Source Import on the repository — see Import IBM i Sources and Generate Build Rules.
3. Add an IBM i connection
On the Connections tab, click Add Connection → IBM i, and enable the features you need:
- SQL — database access via RAS (the
sqlskill). Requires the user password. - Build — compile with
codermakeinto an isolated per-task library. Set the Build Repo and a Build Library Name Prefix. - Interactive Sessions — drive 5250 and Rich Display sessions through Genie. Set PUI Base URL, PUI Render Path (defaults to
/profoundui/genie), PUI Launch Path (defaults to/profoundui/auth/genie), and the user/password.
See IBM i connections → Interactive Sessions for every field and the IBM i-side requirements.
4. What gets configured automatically
Enabling Interactive Sessions wires up the Application Server for you — you don't set a Start Command or Proxy URL by hand. CoderFlow:
- points the Application Server proxy at the Profound UI base URL,
- configures basic auth from the connection,
- adds an
X-Agentic-Task-Libheader carrying the per-task build library, so Genie renders against the task's own compiled changes, and - adds a launch URL named Genie (<connection-name>).
It also imports the ibmi-interactive-session skill so agents can operate the sessions.
Check Profound UI reports reachability, version, HTTPS/certificate trust and CoderFlow integration separately. Guided installation also enables these global settings in the selected instance when its configuration can be safely edited:
SetEnv PUI_ALLOW_AGENTIC_TASK_LIB 1
SetEnv PUI_ALLOW_CODERFLOW_PROXY 1
For an existing instance, use Enable / repair CoderFlow integration…. This explicit setup action explains the changes and possible session interruption before applying them and starting/restarting that instance. Checking alone never changes configuration or restarts services. Unresolved includes or overrides require manual review; see the repair and recovery steps.
A successful product install keeps the Use … for interactive sessions hand-off even if integration has a remaining manual step. Saved directives, activation and the sign-on prerequisites are reported separately. PUISETENV stays unverified until you check the configured profile with a fresh Genie sign-on; CoderFlow does not rewrite profiles or custom initial programs. HTTPS/certificate provisioning also remains an administrator task.
5. Build the environment
On the Build tab, click Build Now. When it succeeds, click Save, then Commit & Push.
Run and test it
- Create a task in this environment. The agent edits source and compiles it into the task's temporary build library with
codermake. - Open the task and use the Genie (<connection-name>) launch URL (or the Testing menu) to open the proxied Profound UI. Drive 5250 or Rich Display sessions against the task's library list — see Staged Tasks and QA Mode.
- Sessions are recorded into Task Visualizations for review, alongside the changed-file diffs.
Notes
- Per-task isolation. The
X-Agentic-Task-Libheader means each task's Genie session reflects that task's own compiled changes, isolated from other tasks. - One per environment. An environment can have at most one IBM i connection with the Build, Sync, or Interactive Sessions feature per scope. See Environments → Connections.
Related
- IBM i connections — full connection reference and IBM i-side setup
- IBM i overview — RAS,
codermake, sync, source import, build rules - Web Application Environments — the in-container serving models
- QA Mode and Task Visualizations