Fediwall#1 of run 33923190220 dies in flowise's MCP registry reconcile:
"FAILED authorizing web-app-prometheus: 400 ... Failed to connect to Custom
MCP server: SSE error: request to http://prometheusmcp:8080/mcp failed,
reason: connect ECONNREFUSED", thirty retries apart. web-app-prometheus sits
in that round's preload closure, so discovery offers it, but the log carries
no web-app-prometheus task at all: the role never ran and its sidecar never
served.
Discovery cannot see that. endpoint_unreachable rejects a provider whose
declaration lacks a port or path, a static check; nothing probes whether the
provider is listening. The platform answers runtime absence with
MCP_RECONCILE_STRICT instead, and 001aeda0b1 states the rule: a rejection
"stays non-fatal in the app pass, where a provider may still be waiting", and
"becomes fatal at the reconciliation stage, where every provider has already
run". sys-svc-mcp-reconcile sets the flag true around its loop and false after;
the group_vars default is false.
flowise's registry task never consulted it, so its own early pass failed the
play on a provider that the later convergence would have reached. It now fails
only when strict, which is exactly the reconciliation stage. The retries stay,
so a provider that appears late in the same pass is still picked up, and the
password and identity tasks above keep failing hard because they address
flowise's own container rather than a foreign one.
Two client roles that never had this problem show the shape: web-app-hermes
and web-app-openclaw call the same file from their own 00_core.yml and carry
no retries/until at all. An earlier attempt in this series deleted the call
from flowise and openwebui as a duplicate of the central stage; fifteen roles
make that call, hermes and openclaw among them, so the duplication is the
convention and that change was withdrawn.
Qdrant ran with no authentication at all: its environment set the HTTP port and disabled telemetry and nothing else, so every container that reached its network held full read, write and delete access to every collection, and the MCP surface's read-only promise rested entirely on the adapter's contract naming only GET routes. Qdrant 1.19 forces authentication on every request once a read-only key is set, so the engine now declares both keys, the adapter presents the read-only one, and the full key goes only to consumers that own data here. A write route is now refused by the engine rather than only by the adapter.
Three defects in the Flowise registry path. Its tool-name check compared against a contract the role always passed as empty, so it could not fail; it now compares against the provider's declared allowlist. Its managed flow pointed a customMCP node at an input that exists on no node at this tag, while the node that does take a registry id is customMcpServerTool, and the one that takes an inline config would have embedded the bearer in the flow; the flow is now an Agentflow whose Tool node calls one named tool with fixed arguments and is executed on every deploy. Its module docstring claimed only SSE providers could be registered, while the pinned MCPToolkit connects Streamable HTTP first and falls back to SSE, so the docstring contradicted the role's own declaration.
Reaching a provider inside the container network needs HTTP_SECURITY_CHECK=false, which replaces the built-in deny list rather than merging into it, so a typo there disables every protection silently while the registry keeps working. The deploy now points a throwaway entry at loopback and at the metadata address and fails unless both are refused.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four roles declared the litellm service and never included the gateway's consumer contract, which is the single place that refuses an empty virtual key and an AI base URL pointing outside the container network. They rendered their AI configuration regardless, so a missing key or a public endpoint reached a green deploy and every prompt left the deployment unauthenticated. Flowise, Hermes, OpenClaw and Open WebUI now include it, each gated on its own enable flag.
A lint derives the consumer set from the repository rather than a list: a litellm block without modes is a consumer, while the gateway and its admin UI declare the service they run and are exempt from the contract they own. The existing rule against addressing a model provider directly scanned only templates, so a role resolving the same variable in vars/ was invisible to it; it now scans both, and a negative control keeps the pattern from going quietly stale.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Twelve callers load their scripts with the relative form
lookup('file', 'name.py'), which resolves against the role's files/
directory. dfe7a7fdd moved those scripts into files/python/ without
updating them, so the MCP provisioning and probes of baserow, checkmk,
flowise, gitlab, homeassistant, jellyfin, n8n, openwebui and pretix, plus
n8n's AI gateway task, all resolved a path that no longer exists.
This is the second time the same relocation left callers behind. After
2aec4326a fixed the php paths, the claim that no stale references remained
came from grepping the worktree, which already carried the corrected paths;
HEAD still held the old ones. The check has to run against the commit.
Verified against HEAD this time: the eleven remaining stale relative lookups
are exactly the files in this bundle, and no other loading form (copy src,
script) carries a stale path.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Loading the backend first left flowise's swarm deploy notification pending,
and the front proxy's own handler flush consumed it, so the stack was never
rolled out and openresty answered 502 for thirty attempts. sys-stk-full
loads both in order and derives the domain and port the role passed by hand.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The reconciler addresses a client with tasks_from: utils/mcp.yml, which
only works when every role offers its MCP work at the same path. The
files were named per role, and on the two agents the work sat inline in
the entry point, where nothing but a full role run could reach it.
The filter docstring pointed at the pre-rename path and now names the
file that exists.
Every MCP role keeps its task tree under tasks/utils/ without numeric
prefixes, so a caller can reach any of them through include_role
tasks_from. utils/roles/mapping.py holds that path as the single source.
Ansible resolves a relative include against the including file's own
directory when that file was reached by path, but against tasks/ when it
was reached through tasks_from. The same file therefore loaded from its
own role and failed from a caller using tasks_from. Every nested include
now names {{ role_path }}/tasks/utils/..., which holds under both bases,
and a lint walks each tasks_from entrypoint transitively to keep it so.
The exec-gating lint learns the same form, otherwise it loses the include
chain and reports inherited IS_STACK_HOST gates as missing.
The expression resolving a provider's stored token appeared eighteen
times. plugins/lookup/mcp_credential.py replaces it, and mcp_probe_owner
disappears from eight roles: it was only ever a step towards the bearer
and no probe read it.
The Hermes and OpenClaw probes stop aborting the play while it is still
running. Roles deploy in dependency order, so a client probing during its
own role can run before a provider exists at all, and failing there
prevents that provider from ever deploying. The verdict moves to the
reconciliation pass, gated on MCP_RECONCILE_STRICT, which the reconciler
sets and group_vars declares so the readers need no default.
The sys-svc-mcp-reconcile role is added here; wiring it into a stage is a
separate change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The published image lies about its version: flowiseai/flowise 3.1.3, 3.1.4 and
latest all ship package 3.1.2, measured in the running container rather than
inferred from the tag. An unpinned global npm install layer is reused between
releases. We build from the declared version instead.
The build installs locally rather than globally because flowise declares node
^24 from 3.1.3 on while resolving better-sqlite3 to 9.6.0, whose binding.gyp
pins -std=c++17 against node 24 headers demanding C++20. That version cannot
build there at all, and an npm override is only honoured for a local install.
Three upstream behaviours were read at the pinned source before being worked
around: cookie authentication is routed only behind an x-request-from header,
the key controller rejects an empty permissions list, and key creation answers
with the whole key list instead of the new key.
The identity password is written straight into the user row because flowise
accepts a password at registration only, and registration works once; a
redeploy carrying a rotated credential otherwise meets an account whose hash
still encodes the previous one.
Server registration refuses anything that does not speak SSE. This flowise
authorizes with an SSE toolkit, so registering another transport would claim a
connection that cannot be made.
Register every discovered server as an Open WebUI tool server, skipping
the auth schemes its ToolServerConnection cannot present so none is
stored with an unusable credential. Serialise that value with compact
separators: a space makes dotenv_quote wrap it, swarm keeps the quotes
inside the value, and Open WebUI then parses a string instead of a list
and crash-loops on startup.
Prepare the Flowise instance instead of registering servers in it.
Flowise 3.1.3 exposes no instance-level MCP registry; MCP lives in a
per-flow Custom MCP node. The role pins the transport to URL-based
servers so no flow can spawn a local command, and relaxes the host deny
list far enough to reach a container origin while loopback and the
cloud-metadata addresses stay denied. Selecting a server inside a flow
stays an operator step, and the README says so.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Flowise shipped its own LiteLLM container: own image, own port, own
config template, own volume, own master key. Every further consumer of
the model plane would have needed the same again.
The service entry becomes a plain consumer flag on svc-ai-litellm, so the
include resolver pulls the gateway into the round and the flag is false
when no gateway is deployed. The sidecar render task, its template, its
volume and the compose_handlers_flush dance that existed only to sequence
that render are all gone; 00_core.yml is back to the standard shape.
Two corrections ride along in env.j2. The endpoint keeps its /v1 suffix,
which the OpenAI-compatible surface requires. And Flowise now presents
its own virtual key from credentials.litellm_api_key rather than the
gateway master key, which previously handed a consumer full
administrative access to the gateway.
The whole block is conditional now, so without a gateway Flowise renders
no OPENAI_API_BASE at all instead of an address pointing nowhere.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add tests/lint/repository/test_task_file_numbering.py: numbered task
files use exactly two leading digits, a numbered *_core.yml must be
00_core.yml, per directory the numbers are unique/consecutive and start
at 01 (00 with a core file), numbered includes appear in ascending
order, and shared helpers live in tasks/utils/ instead of the numbered
chain.
Apply it across every role: rename NN_core.yml to 00_core.yml with a
consecutive down-shift of the siblings, move letter-suffix children into
a subdirectory named after their parent task, relocate helpers included
from more than one place into tasks/utils/, and update every reference
(include_tasks, path_absolute, .ansible-lint-ignore, requirement docs).
Legacy narration comments in the touched files are deleted; the upstream
xwiki.properties reference block opts out via nocheck. The unused
MAILU_IP6_PUBLIC var (only ever named in now-removed comments) is dropped.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Introduce plugins/lookup/path_absolute.py as the single source for the
repo-root anchor: lookup('path_absolute', 'roles/...') returns an absolute
path under playbook_dir, joining every term. Replace all hand-built
playbook_dir path constructions across .yml/.yaml and .j2 templates with it:
[playbook_dir, '<rel>'] | path_join -> lookup('path_absolute', '<rel>')
playbook_dir ~ '/<rel>' -> lookup('path_absolute', '<rel>')
{{ playbook_dir }}/<rel> -> lookup('path_absolute', '<rel>')
Unrelated repo_path symbols in cli/meta/roles/applications/resolution and the
single bare discover_playwright_roles(playbook_dir) call are left untouched.
Add tests/lint/ansible/test_playbook_dir_via_lookup.py: fails the build if any
of the three construction forms reappears in .yml/.yaml/.j2, pointing callers
at the lookup. Unit test covers the plugin.
Touching these files pulled their pre-existing narration comments into
test_comments_valid scope; clean them (delete restatement/banners, keep real
trip-wires as Workaround/Upstream/Invariant markers). Drop caution, important
and pitfall from the comment-lint marker vocabulary and remove the comments
that used them.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add tests/lint/repository/test_name_emoji_and_length.py: every Ansible task
name and GitHub Actions workflow/job/step name must start with an emoji (first
char not 0-9/A-Za-z) and be <= 80 chars.
- emit emoji names at the source: cli/build/include generator now writes
"🔧 setup <app>" / "🚿 flush handlers after <app>" so generated
tasks/groups/*-roles.yml satisfy the convention too
- convert existing task/step/job/workflow names to emoji-prefixed, choosing a
task-fitting icon per action (📦 install, ✅ ensure, 🔍 probe, 🥾 bootstrap,
🚀 deploy, 🧩 service-module load, ⬇️ checkout, 📂 include-routines, ...)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Resolve 69 conflicts across ~70 files: union main's image-version bumps and new service fields (bond, resource limits) with the swarm branch's default_placement / NFS / IS_STACK_HOST config; retire the three per-deploy-type test workflows main still edited (consolidated into compose+swarm); adopt main's run-once 01_core dispatchers while re-applying each role's swarm guard. Fix the post-merge lint surfaced by main's stricter rules: bond + resource-budget keys, container-exec stack-host gating + resolver routing, leaky/dead proxy vars, suitecrm once-flag, xwiki variant nocheck, nextcloud oversized-block to gated include, matrix-mdad capped at 3g. make autoformat clean, make test green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
test_service_core_first_task_run_once requires tasks/01_core.yml to set the
run_once flag (include_tasks: utils/once/flag.yml) as its very first task so the
guard is in place before the role body runs and the variant-aware loader cannot
recurse. Move the flag include from the end to the top in the shared-provider
roles that had it last (bigbluebutton, flowise, gitlab, matrix, mattermost,
moodle, openproject, openwebui, peertube, suitecrm, xwiki).
Also give the template-only providers svc-db-redis and svc-db-memcached the
run-once guard (main.yml includes 01_core.yml when run_once_<role> is undefined;
01_core.yml sets the flag) so they satisfy the shared-provider run_once guard.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Declare the 14 services that consuming roles reference but no role provided
(bigbluebutton, xwiki, openproject, gitlab, mattermost, matrix, zammad,
openwebui, flowise, peertube, moodle, suitecrm, redis, memcached) as shared
providers: set enabled/shared/bond on each entity so the service registry
registers them and the dangling references resolve.
Add test_shared_service_provider.py (lint): fail when a referenced shared
service has no provider and its name matches a role's entity/provider that does
not register it (suppressable via '# nocheck: shared-provider').
Guard the 9 web-app shared services with the run-once 01_core.yml pattern
(main.yml includes 01_core.yml only when run_once_<role> is undefined; the core
flags itself via utils/once/flag.yml) so they deploy once when pulled in by
multiple consumers. svc-db-redis/memcached are template-only (no runtime tasks).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pure 1:1 rename (69 insertions / 69 deletions across 66 files) to
finish the ``docker_*`` → ``compose_*`` alignment started in
8373468e9. The flag configures whether the compose include flushes
handlers right after run; the ``docker_`` prefix was misleading
(handlers are Ansible-side, not docker-side) and the variable is part
of the same ``sys-svc-compose`` SPOT cluster as ``compose_repository_*``.
Touches the role default, the implementation
(``sys-svc-compose/tasks/utils/up.yml``), all 60+ consumer roles,
the network-utility wrapper + its README, and the
``test_network_create_via_util.py`` lint (both the docstring guidance
and the asserted diff).
Breaking change for any external override that still sets the old
name — no alias is provided.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* scripts/lint/autoformat.sh:
- rename run_sort_claude_settings worker to run_claude (matches the
renamed scripts/lint/claude.sh).
- drop the EXIT trap that wiped status_dir; cleanup happens explicitly
after `wait` returns, so a parallel worker that interrupts the
parent never races a vanished status dir.
- each worker now captures its tool's exit code and writes one of
OK / SKIP / FAIL. Tolerated `expected violations remain`-style exits
(markdownlint-cli2 rc≤1, ansible-lint rc≤2) still count as OK; only
real failures count as FAIL.
- all workers run to completion regardless of individual tool failures
(so the user always sees the full picture); after the summary, the
script exits 1 if any worker reported FAIL.
* scripts/lint/sort_claude_settings.sh -> scripts/lint/claude.sh
(history preserved via git mv); update test_claude_settings_sorted.py
fix-hint accordingly.
Plus working-tree drift across roles/, group_vars/, inventories/,
requirements/, and .pre-commit-config.yaml accumulated from the
make-test setup phase regenerating generated artefacts during recent
test runs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single atomic landing for three back-to-back role-meta refactor requirements
(see docs/requirements/008-role-meta-layout.md, 009-per-role-networks-and-ports.md,
010-role-meta-runafter-lifecycle-migration.md).
req-008 — role meta layout
* Move every role's metadata under meta/<topic>.yml with file-root
convention (no wrapping key matching the basename):
- schema/main.yml + config/main.yml.credentials → meta/schema.yml
- users/main.yml → meta/users.yml
- config/main.yml.server → meta/server.yml
- config/main.yml.rbac → meta/rbac.yml
- config/main.yml.compose.services + every other top-level key
of config/main.yml inlined under <entity>.<key>
→ meta/services.yml
- config/main.yml.compose.volumes → meta/volumes.yml
* Drop the legacy schema/, users/, config/ directories from every role.
* meta/schema.yml: support nested credential keys, optional `default:`
field (Jinja literal, not rendered/validated/regenerated), implicit
`algorithm: plain`.
* Materialised paths drop the `compose.` prefix:
compose.services.<e>.<…> → services.<e>.<…>
compose.volumes.<k> → volumes.<k>
<former-top-level>.<…> → services.<entity_name>.<former-top-level>.<…>
* Update every Python consumer (utils/cache, utils/manager, utils/applications,
plugins/lookup, plugins/filter, cli/*) plus every Jinja template, vars/*.yml
and task file to the new paths.
* No legacy fallback. Repository-wide grep for `compose.services.` /
`compose.volumes.` / `schema/main.yml` / `users/main.yml` / `config/main.yml`
is empty outside docs/requirements/ and tasks/utils/migrate_meta_layout.py.
* Add tests/lint/repository/test_role_meta_layout.py guarding against the
old paths returning, and an explicit lint guard that meta/variants.yml
files do not carry a `compose:` wrapper.
req-009 — per-role networks/ports
* Move group_vars/all/08_networks.yml.defaults_networks.local.<role>.{subnet,
dns_resolver} into roles/<role>/meta/server.yml.networks.local.{subnet,
dns_resolver}.
* Delete group_vars/all/09_ports.yml entirely; ports now live on the
service entity that exposes them at
roles/<role>/meta/services.yml.<entity>.ports.{inter,local,public}.
* Rename + int-cast every former top-level `services.<entity>.port`
field to `services.<entity>.ports.inter` (single int).
* Always-category-keyed maps for ports.local / ports.public; relay
spans use {start, end} integers.
* Add PORT_BANDS map to group_vars/all/08_networks.yml — single source
of truth for the per-<scope>.<category> band ranges, consumed by both
the new CLI helpers and the lint test.
* New CLI helpers (cli meta ports suggest, cli meta networks suggest)
walk the live role tree gap-first then increment-fallback, with
deterministic output and non-zero exit on capacity overflow.
* cli create role consumes both helpers when scaffolding a new role.
* Lint extensions: bare-int port shape rejection, host-bound port
collision check (single ints + relay span integers, inter excluded),
band-membership check with documented allow-list for the legacy
web-app-bigbluebutton.bigbluebutton.local.http: 48087 entry.
req-010 — run_after / lifecycle migration
* Move galaxy_info.run_after and galaxy_info.lifecycle out of
meta/main.yml into meta/services.yml.<primary_entity>.{run_after,
lifecycle}; <primary_entity> is the value returned by
get_entity_name(role_name).
* For multi-entity roles whose primary entity has no compose service
(web-app-bluesky → bluesky, web-app-matrix → matrix, desk/dev/drv/
gen roles), create a dedicated metadata-only top-level entry under
<primary_entity>.
* Introduce utils/roles/meta_lookup.py with get_role_run_after /
get_role_lifecycle that resolve <primary_entity> once and read from
meta/services.yml; degrade gracefully when meta/services.yml is
absent OR the field is absent; raise MetaServicesShapeError when the
file exists but is malformed.
* Update every consumer (utils/service_registry, utils/invokable,
cli/meta/applications/resolution/run_after/__main__, cli/meta/roles/
lifecycle_filter/__main__, cli/meta/applications/resolution/combined,
cli/build/{tree,graph,role_include}, cli/deploy/development/{common,
init,deps,deploy}, plugins/filter/canonical_domains_map, etc.) to
use the helper.
* Lint guards: no run_after/lifecycle in meta/main.yml; lifecycle
values restricted to {planned,pre-alpha,alpha,beta,stable,deprecated};
run_after/lifecycle never on a non-primary entity; empty run_after
lists rejected.
Also folded into this change set
* services_disabler now writes `services.<svc>.{enabled,shared}` into
host_vars instead of the old `compose.services.<svc>.{...}` path; the
discover/find pair walks meta/services.yml and only treats entries
carrying enabled/shared/provides as provider declarations (so
metadata-only entries from req-010 do not create spurious provider
rows).
* service_registry skips `run_after` deps whose target was filtered
out via SERVICES_DISABLED instead of failing the whole load.
* cli/mirror/resolver and cli/create/inventory/mirror_overrides emit/
consume the new applications.<role>.services.<svc>.image shape
(mirror.md updated to match).
* utils/domains/list.build_applications_from_roles now reads
meta/server.yml directly (it had been silently returning {} since
config/main.yml stopped existing).
* utils/roles/resource_validation.filter_roles_by_min_storage reads
meta/services.yml.<entity>.min_storage instead of the legacy
compose.services path.
* web-app-bluesky pre-clones the social-app build context BEFORE
sys-svc-compose so the build context is on disk by the time
sys-stk-front-proxy/tasks/main.yml flushes the docker build handler.
* 4 meta/variants.yml files (svc-ai-ollama, web-app-phpmyadmin,
web-app-wordpress, web-app-keycloak) lose their `compose: services:`
wrapper so deep-merge over meta/services.yml lands on the new
file-root shape.
Validation
* Unit tests: 1571 passed.
* Lint tests: 50 passed.
* Integration tests: 161 passed (1 skipped).
* Full-cycle fresh-purged-apps deploy succeeded for the 8-app validation
set: web-app-{nextcloud,wordpress,yourls,bigbluebutton,bluesky,gitea,
discourse,mattermost}.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Switch all docker compose invocations to the compose wrapper
- Update comments, task names, and log messages accordingly
- Simplify handlers and tasks to rely on wrapper-managed flags and env files
- Align tests and documentation with compose-based workflow
https://chatgpt.com/share/69866ede-be2c-800f-94b0-bffcf8fed260
Unify reverse proxy handling across apps via sys-stk-front-proxy and cleanly separate internal docker nginx configs from external vhosts.
Changes:
- docker-compose: use with_first_found for optional env and docker-compose.override files so roles can provide either a template or a static file without noisy 'Could not find or access' failures.
- sys-stk-front-proxy: remove front_proxy_domain_conf_src and switch nginx vhost rendering to with_first_found over app-specific proxy.conf.j2 and the generic roles/sys-svc-proxy vhost flavour; keep health-check and handler logic unchanged.
- web-app-nextcloud: migrate to sys-stk-full-stateful (front proxy + DB + docker), move internal nginx config to docker.conf.j2 under the volume path, and rename host.conf.j2 to proxy.conf.j2 for the external vhost.
- web-app-magento: rename nginx.conf.j2 to docker.conf.j2 and update the runtime template task accordingly to make the intent (internal nginx) explicit.
- web-app-matrix: rename nginx.conf.j2 to synapse.conf.j2 and adjust the webserver task to use the new template name for the synapse vhost.
- web-app-bridgy-fed & web-app-flowise: pass domain and http_port explicitly when including sys-stk-front-proxy so the front stack has all required context.
- web-svc-cdn/file/html: replace direct sys-stk-front-base + sys-util-csp-cert + nginx.conf.j2 handling with sys-stk-front-proxy and proxy.conf.j2, relying on the shared front-stack for TLS/CSP/vHost wiring.
- web-svc-collabora: drop the direct nginx.conf.j2 vhost generation and rename it to proxy.conf.j2 so it is picked up by sys-stk-front-proxy like other services.
- web-opt-rdr-domains: rename redirect.domain.nginx.conf.j2 to redirect-domain.conf.j2 and adjust the task for clearer and more consistent naming.
Context: see ChatGPT refactor discussion on 2025-11-30 (proxy unification, Collabora/Nextcloud/CDN stacks, CSP/header handling): https://chatgpt.com/share/692c64ea-a488-800f-ad42-7f7692a3742f
• 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
