This update waits for the Keycloak container to become healthy before attempting login and replaces the old realm-management based role assignment with the global 'admin' realm role.
See: https://chatgpt.com/share/68e99953-e988-800f-8b82-9ffb14c11910
• Add svc-ai category under roles and load it in constructor stage
• Create new 'svc-ai-ollama' role (vars, tasks, compose, meta, README) and dedicated network
• Refactor former AI stack into separate app roles: web-app-flowise and web-app-openwebui
• Add web-app-minio role; adjust config (no central DB), meta (fa-database, run_after), compose networks include, volume key
• Provide user-focused READMEs for Flowise, OpenWebUI, MinIO, Ollama
• Networks: add subnets for web-app-openwebui, web-app-flowise, web-app-minio; rename web-app-ai → svc-ai-ollama
• Ports: rename ai_* keys to web-app-openwebui / web-app-flowise; keep minio_api/minio_console
• Add group_vars/all/17_ai.yml (OLLAMA_BASE_LOCAL_URL, OLLAMA_LOCAL_ENABLED)
• Replace hardcoded include paths with path_join in multiple roles (svc-db-postgres, sys-service, sys-stk-front-proxy, sys-stk-full-stateful, sys-svc-webserver, web-svc-cdn, web-app-keycloak)
• Remove obsolete web-app-ai templates/vars/env; split Flowise into its own role
• Minor config cleanups (CSP flags to {}, central_database=false)
https://chatgpt.com/share/68d15cb8-cf18-800f-b853-78962f751f81
- Move all domain→expected-status mapping to filter `web_health_expectations`.
- Require explicit app selection via non-empty `group_names`; only those apps are included.
- Add `www_enabled` flag (wired via `WWW_REDIRECT_ENABLED`) to generate/force www.* → 301.
- Support `redirect_maps` to include manual redirects (sources forced to 301), independent of app selection.
- Aliases always 301; canonicals use per-key override or `server.status_codes.default`, else [200,302,301].
- Remove legacy fallbacks (`server.status_codes.home` / `landingpage`).
- Wire filter output into systemd ExecStart script as JSON expectations.
- Normalize various templates to use `to_json` and minor spacing fixes.
- Update app configs (e.g., YOURLS default=301; Confluence default=302; Bluesky web=405; MediaWiki/Confluence canonical/aliases).
- Constructor now uses `WWW_REDIRECT_ENABLED` for domain generation.
Tests:
- Add comprehensive unit tests for filter: selection by group, keyed/default codes, aliases, www handling, redirect_maps, input sanitization.
- Add unit tests for the standalone checker script (JSON parsing, OK/mismatch counting, sanitization).
See conversation: https://chatgpt.com/share/68c2b93e-de58-800f-8c16-ea05755ba776
- Converted group_vars/all/13_ldap.yml from lower-case to ALL-CAPS nested keys.
- Updated all roles, tasks, templates, and filter_plugins to reference LDAP.* instead of ldap.*.
- Fixed Keycloak JSON templates to properly quote Jinja variables.
- Adjusted svc-db-openldap filter plugins and unit tests to handle new LDAP structure.
- Updated integration test to only check uniqueness of TOP-LEVEL ALL-CAPS constants, ignoring nested keys.
See: https://chatgpt.com/share/68b01017-efe0-800f-a508-7d7e2f1c8c8d
- Allow kc_object_kind='realm'
- Map endpoint to 'realms' and default lookup_field to 'id'
- Use realm-specific kcadm GET/UPDATE (no -r flag)
- Preserve immutables: id, realm
- Guard query-based ID resolution to non-realm objects
Context: fixing failure in 'Update REALM mail settings' task.
See: https://chatgpt.com/share/68affdb8-3d28-800f-8480-aa6a74000bf8
- Fixed incorrect filter usage in docker-compose handler (proper use of | path_join).
- Improved LetsEncrypt template by joining paths with filenames instead of appending manually.
- Enhanced sys-svc-msmtp task with an additional condition to only run if no-reply mailu_token exists.
- Updated Keycloak meta to depend on Mailu (ensuring token generation before setup).
- Refactored Keycloak import path variables to use path_join consistently.
- Adjusted Mailu meta dependency to run after Matomo instead of Keycloak.
See: https://chatgpt.com/share/68af13e6-edc0-800f-b76a-a5f427837173
- Extended kc_force_attrs in tasks/main.yml to source 'publicClient',
'serviceAccountsEnabled' and 'frontchannelLogout' directly from
KEYCLOAK_DICTIONARY_REALM for consistency with import definitions.
- Updated default.json.j2 import template to set 'publicClient' to true.
- Public client mode is required so the frontend API of role 'web-app-desktop'
can handle login/logout flows without client secret.
Ref: https://chatgpt.com/share/68ae0060-4fac-800f-9f02-22592a4087d3
- Render userObjectClasses via `tojson` (and trim) to avoid invalid control
characters and ensure valid realm import parsing.
- Introduce KEYCLOAK_LDAP_USER_OBJECT_CLASSES in vars; exclude `posixAccount`
for Keycloak’s LDAP config while keeping it for Ansible-managed UNIX users.
- Update UserStorageProvider template to use the new variable.
Rationale:
Keycloak must not require `posixAccount` on every LDAP user. We keep
`posixAccount` structural for Ansible provisioning, but filter it out for
Keycloak to prevent sync/import errors on entries without POSIX attributes.
Touched:
- roles/web-app-keycloak/templates/import/components/org.keycloak.storage.UserStorageProvider.json.j2
- roles/web-app-keycloak/vars/main.yml
Refs: conversation https://chatgpt.com/share/68aa1ef0-3658-800f-bdf4-5b57131d03b4
- Improved get_service_name filter plugin (clearer suffix handling, consistent var names).
- Added MODE_ASSERT flag to optionally execute validation/assertion tasks.
- Fixed systemd unit handling: consistent use of %I instead of %i, correct escaping of instance names.
- Unified on_failure behavior and alarm composer scripts.
- Cleaned up redundant logging, handlers, and debug config.
- Strengthened sys-service template resolution with assert (only active when MODE_ASSERT).
- Simplified timer and suffix handling with get_service_name filter.
- Hardened sensitive tasks with no_log.
- Added conditional asserts across roles (Keycloak, DNS, Mailu, Discourse, etc.).
These changes improve consistency, safety, and validation across the automation stack.
Conversation: https://chatgpt.com/share/68a4ae28-483c-800f-b2f7-f64c7124c274
- Fixed incorrect handling of RBAC group configuration (moved from OIDC claims into dedicated RBAC variable set).
- Unified RBAC group usage across applications (LAM, pgAdmin, phpLDAPadmin, phpMyAdmin, YOURLS).
- Replaced old 'KEYCLOAK_OIDC_RBAC_SCOPE_NAME' with dedicated 'KEYCLOAK_RBAC_GROUP_*' variables.
- Updated OAuth2 Proxy configuration to use 'RBAC.GROUP.CLAIM'.
- Refactored Keycloak role task structure:
* Renamed and reorganized task files for clarity ('_update.yml', '02_cleanup.yml', etc.).
* Introduced meta and dependency handling separation.
- Cleaned up Keycloak config defaults and recaptcha placeholders.
Replace 01_import.yml with 01_initialize.yml (KEYCLOAK_HOST_IMPORT_DIR)
Add generic 02_update.yml (kcadm updater for clients/components)
- Resolve ID → read current → merge (kc_merge_path optional)
- Preserve immutable fields; support kc_force_attrs
Update tasks/main.yml:
- Readiness via KEYCLOAK_MASTER_REALM_URL; kcadm login
- Merge LDAP component config from Realm when KEYCLOAK_LDAP_ENABLED
- Update client settings incl. frontchannel.logout.url
realm.json.j2: include ldap.json in UserStorageProvider
ldap.json.j2: use KEYCLOAK_LDAP_* vars for bindDn/credential/connectionUrl
vars/main.yml: add KEYCLOAK_* URLs/dirs and KEYCLOAK_DICTIONARY_REALM(_RAW)
docker-compose.yml.j2: mount KEYCLOAK_HOST_IMPORT_DIR
Cleanup: remove 02_update_client_redirects.yml, 03_update-ldap-bind.yml, 04_ssh_public_key.yml; drop obsolete config flag; formatting
Note: redirectUris/webOrigins ordering may still cause changed=true; consider sorting for stability in a follow-up.
- Removed all static 'id' fields from realm.json.j2, ldap.json.j2, and client.json.j2
- Replaced 'desktop-secret' with correct 'client-secret' authenticator type
- Standardized Jinja filters to use 'to_json' consistently
- Corrected defaultClientScopes entry from 'web-app-origins' to built-in 'web-origins'
- Verified LDAP mapper definitions and optional realm role mapping
- Ensured realm.json.j2 contains only required scopes
References: Chat with ChatGPT (2025-08-17)
https://chatgpt.com/share/68a1aaae-1b04-800f-aa8d-8a0ef6d33cba
- Replace KEYCLOAK_KCADM_PATH with KEYCLOAK_EXEC_KCADM consistently
- Externalize client.json to separate Jinja2 template and include it in realm.json
- Simplify LDAP bind update to use explicit KEYCLOAK_LDAP_* vars
- Add async/poll support for long-running kcadm updates
- Restructure vars/main.yml: clearer grouping (General, Docker, Server, Update, LDAP, API)
- Compute redirectUris/webOrigins centrally in vars
- Align post.logout.redirect.uris handling with playbook
Conversation: https://chatgpt.com/share/68a1a11f-f8ac-800f-bada-cdc99a4fa1bf
Switch to command:argv to avoid shell quoting and argument splitting issues.
Pass -s config values as JSON arrays via to_json, fixing previous errors: Cannot parse the JSON / failed at splitting arguments.
Also reconcile config.connectionUrl from ldap.server.uri.
Source desired values strictly from ldap.* (no computed defaults) and assert their presence.
Keep operation idempotent by reading current values and updating only on change.
Minor refactor: build reusable kcadm_argv_base and expand client state extraction.
Touch: roles/web-app-keycloak/tasks/03_update-ldap-bind.yml
https://chatgpt.com/share/689bea84-7188-800f-ba51-830a0735f24c
