Files
sure/test/controllers/admin/system_health_controller_test.rb
T
Juan José MataandClaude f78303ebbe Add function-calling probe to detect tool-use support (#3255)
* feat(ai-health): name the missing function calling behind an opaque chat error

The assistant reads accounts, transactions and holdings through function
calls, so every chat request carries a `tools` payload. A model without
function-calling support rejects it — OpenRouter answers a bare 404 — and
the operator sees only that status code, with nothing pointing at the
model. Both earlier attempts at this guessed from the chat-time error;
the AI status page already runs live probes, so let it answer the
question directly instead.

`AiHealth::Probe#function_calling` asks the configured model for one
trivial tool call the way the assistant asks for its own: chat
completions with `tools` for OpenAI-compatible endpoints, the Responses
API for hosted OpenAI, and `messages.create` with `tools` for Anthropic,
carrying the same strict schema `Provider::Openai` sends. Reading it
against the plain LLM probe is what makes the verdict sound rather than
a guess at 404s: plain chat passing while the same request with tools
fails means the model has no function calling; a response that carries
no tool call means the endpoint took the tools but the model ignored
them; both failing, or a timeout, stays an ordinary probe failure.

The AI status card gains a Function calling (tools) row, an alert
naming the fix for each of the two bad outcomes, and a failure reason.
The hosting settings model field now says the assistant needs a
tools-capable model and links super admins to the check.

Refs #830

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DWCmRQ1ry26JnhA79s1ZKH

* fix(ai-health): only call a refusal a refusal, and probe the route chat takes

Two findings from review of the function-calling probe.

A tools request can fail for reasons that say nothing about tool support:
a 429, a 500, a dropped connection, an unreadable body. Reading any
non-timeout failure as `:unsupported` sent the operator hunting for a new
model over a transient blip. Only a 4xx the service chose to answer with
— excluding the ones that mean "not now" or "not you" — is a refusal of
the tools payload; everything else stays an ordinary probe failure. The
bare 404 from OpenRouter that this page exists to explain still reads as
missing function calling.

`Provider::Openai#supports_responses_endpoint?` is the real routing
decision and `OPENAI_SUPPORTS_RESPONSES_ENDPOINT` can flip it either way,
so choosing the API from "is the endpoint custom" could probe Chat
Completions while chat uses Responses, or the reverse — reporting on a
path the assistant never takes. Ask the provider instead, and cache the
two routes under separate keys.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DWCmRQ1ry26JnhA79s1ZKH

* fix(ai-health): confirm a tools refusal before blaming the model

A client error on the tools request can mean "your tools payload" or "your
request, tools or not" — an invalid schema, a route the endpoint does not
serve, a model it will not run. Splitting those on the status code alone
still put a 422 from an endpoint contract on the model's account and told
the operator to go find another one.

The probe now confirms it: when the tools request comes back a client
error, it asks again with the tools taken off. Only if that lands is the
tools payload what was turned down, and the probe says so with its own
failure code — provider-agnostic, and no reading of error text for the
word "tool", which would only ever fit the provider it was written
against. Statuses that mean "not now" or "not you" (401, 402, 403, 408,
429) never get a second ask. `AiHealth` now just reports the probe's
verdict instead of inferring one from the status.

The troubleshooting fix no longer points at an OpenRouter free tier:
those providers commonly log prompts and completions for training, and
every assistant tool call carries accounts, transactions, and holdings.
It points at the model recommendations already in this doc, and says why
free tiers are the wrong place to look.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DWCmRQ1ry26JnhA79s1ZKH

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-08-31 19:47:48 +02:00

354 lines
14 KiB
Ruby

require "test_helper"
class Admin::SystemHealthControllerTest < ActionDispatch::IntegrationTest
AI_ENVIRONMENT = %w[
OPENAI_ACCESS_TOKEN OPENAI_URI_BASE OPENAI_MODEL OPENAI_REQUEST_TIMEOUT
OPENAI_SUPPORTS_PDF_PROCESSING OPENAI_SUPPORTS_RESPONSES_ENDPOINT
ANTHROPIC_ACCESS_TOKEN ANTHROPIC_API_KEY
ANTHROPIC_BASE_URL ANTHROPIC_MODEL ANTHROPIC_REQUEST_TIMEOUT
VECTOR_STORE_PROVIDER EMBEDDING_URI_BASE EMBEDDING_MODEL
EMBEDDING_DIMENSIONS EMBEDDING_ACCESS_TOKEN QDRANT_URL QDRANT_API_KEY
AI_HEALTH_PROBE_TIMEOUT AI_HEALTH_PROBE_CACHE_TTL
].index_with(nil).freeze
setup do
Setting.stubs(:llm_provider).returns("openai")
Setting.stubs(:openai_access_token).returns(nil)
Setting.stubs(:openai_uri_base).returns(nil)
Setting.stubs(:openai_model).returns(nil)
Setting.stubs(:anthropic_access_token).returns(nil)
Setting.stubs(:anthropic_base_url).returns(nil)
Setting.stubs(:anthropic_model).returns(nil)
AiHealth::Probe.any_instance.stubs(:llm).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:function_calling).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:pdf_text_extraction).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:pdf_vision_processing).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:openai_vector_store).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:pgvector).returns(probe_result(:passing))
AiHealth::Probe.any_instance.stubs(:embedding).returns(probe_result(:passing))
end
test "super admin can view the system health page" do
sign_in users(:sure_support_staff)
SidekiqHealth.any_instance.stubs(:healthy?).returns(true)
SidekiqHealth.any_instance.stubs(:processes_count).returns(1)
SidekiqHealth.any_instance.stubs(:last_heartbeat_at).returns(Time.current)
SidekiqHealth.any_instance.stubs(:max_queue_latency).returns(0.0)
SidekiqHealth.any_instance.stubs(:enqueued_count).returns(0)
SidekiqHealth.any_instance.stubs(:retry_count).returns(0)
SidekiqHealth.any_instance.stubs(:failed_count).returns(0)
SidekiqHealth.any_instance.stubs(:processed_count).returns(42)
SidekiqHealth.any_instance.stubs(:queue_breakdown).returns([ [ "default", 0, 0.0 ] ])
get admin_system_health_url
assert_response :success
assert_match(/Sidekiq status/, response.body)
assert_match(/Healthy/, response.body)
assert_select "button[role='tab']", text: "AI status"
assert_select "[data-ds--tabs-navigate-on-change-value='true']"
end
test "renders degraded state with reason when Sidekiq is unhealthy" do
sign_in users(:sure_support_staff)
SidekiqHealth.any_instance.stubs(:healthy?).returns(false)
SidekiqHealth.any_instance.stubs(:reason).returns(:no_worker_processes)
SidekiqHealth.any_instance.stubs(:processes_count).returns(0)
SidekiqHealth.any_instance.stubs(:last_heartbeat_at).returns(nil)
SidekiqHealth.any_instance.stubs(:max_queue_latency).returns(0.0)
SidekiqHealth.any_instance.stubs(:enqueued_count).returns(7)
SidekiqHealth.any_instance.stubs(:retry_count).returns(0)
SidekiqHealth.any_instance.stubs(:failed_count).returns(0)
SidekiqHealth.any_instance.stubs(:processed_count).returns(0)
SidekiqHealth.any_instance.stubs(:queue_breakdown).returns([])
get admin_system_health_url
assert_response :success
assert_match(/Degraded/, response.body)
assert_match(/No Sidekiq worker process is connected/, response.body)
end
test "non super admin is redirected away" do
sign_in users(:family_admin)
get admin_system_health_url
assert_redirected_to root_path
end
test "unauthenticated user is redirected to sign in" do
get admin_system_health_url
assert_redirected_to new_session_path
end
test "AI status reports the default OpenAI LLM and hosted vector store" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
with_ai_environment("OPENAI_ACCESS_TOKEN" => "sk-secret-openai") do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_select "button[role='tab'][aria-selected='true']", text: "AI status"
assert_match(/LLM and PDF processing/, response.body)
assert_select "[data-testid='selected-llm-provider']", text: "OpenAI"
assert_select "[data-testid='effective-llm-provider']", text: "OpenAI"
assert_match(/gpt-4\.1/, response.body)
assert_match(%r{https://api\.openai\.com/v1}, response.body)
assert_match(/OpenAI hosted vector store/, response.body)
assert_match(/Live check passed/, response.body)
assert_match(/Live checks passed/, response.body)
assert_match(/PDF text-extraction path/, response.body)
assert_match(/PDF vision\/native path/, response.body)
assert_equal 2, response.body.scan(/Synthetic PDF check passed/).size
assert_no_match(/sk-secret-openai/, response.body)
end
test "background jobs tab does not run AI probes" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
AiHealth::Probe.any_instance.expects(:llm).never
AiHealth::Probe.any_instance.expects(:function_calling).never
AiHealth::Probe.any_instance.expects(:pdf_text_extraction).never
AiHealth::Probe.any_instance.expects(:pdf_vision_processing).never
AiHealth::Probe.any_instance.expects(:openai_vector_store).never
with_ai_environment("OPENAI_ACCESS_TOKEN" => "sk-secret-openai") do
get admin_system_health_url
end
assert_response :success
assert_match(/Not checked/, response.body)
assert_no_match(/sk-secret-openai/, response.body)
end
test "AI status warns when a custom OpenAI endpoint is paired with the hosted vector store" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
with_ai_environment(
"OPENAI_ACCESS_TOKEN" => "local-token",
"OPENAI_URI_BASE" => credentialed_url(
scheme: "http",
host: "ollama",
port: 11_434,
path: "/v1",
user: "operator",
password: "uri-secret",
query: "api_key=query-secret"
),
"OPENAI_MODEL" => "qwen3:8b"
) do
AiHealth::Probe.any_instance.stubs(:openai_vector_store).returns(
probe_result(:failing, failure_code: :request_failed, http_status: 404)
)
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_select "[data-testid='selected-llm-provider']", text: "OpenAI-compatible"
assert_select "[data-testid='effective-llm-provider']", text: "Ollama"
assert_match(/OpenAI-compatible API credentials/, response.body)
assert_match(%r{http://ollama:11434/v1}, response.body)
assert_match(%r{did not pass the /v1/vector_stores liveness check}, response.body)
assert_match(/use pgvector with a separate embeddings endpoint/, response.body)
assert_match(/Live check failed/, response.body)
assert_no_match(/local-token|uri-secret|query-secret/, response.body)
end
test "AI status names the missing function-calling support behind an unhelpful chat error" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
AiHealth::Probe.any_instance.stubs(:function_calling).returns(
probe_result(:failing, failure_code: :tools_refused, http_status: 404)
)
with_ai_environment(
"OPENAI_ACCESS_TOKEN" => "router-secret",
"OPENAI_URI_BASE" => "https://openrouter.ai/api/v1",
"OPENAI_MODEL" => "tngtech/deepseek-r1t2-chimera:free"
) do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_select "[data-testid='function-calling-status']", text: /Not supported by the effective provider/
assert_match(/The model does not support function calling/, response.body)
assert_match(/Function-calling failure reason/, response.body)
assert_no_match(/router-secret/, response.body)
end
test "AI status separates a model that ignores tools from one that cannot use them" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
AiHealth::Probe.any_instance.stubs(:function_calling).returns(
probe_result(:failing, failure_code: :no_tool_call)
)
with_ai_environment("OPENAI_ACCESS_TOKEN" => "sk-secret-openai") do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_select "[data-testid='function-calling-status']", text: /Tools accepted, but the model called none/
assert_match(/answered without calling the tool it was asked to call/, response.body)
assert_no_match(/The model does not support function calling/, response.body)
end
test "AI status reports text and vision PDF probes separately" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
AiHealth::Probe.any_instance.stubs(:pdf_vision_processing).returns(
probe_result(:failing, failure_code: :invalid_response)
)
with_ai_environment("OPENAI_ACCESS_TOKEN" => "sk-secret-openai") do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_match(/PDF text-extraction path/, response.body)
assert_match(/PDF vision\/native path/, response.body)
assert_match(/The synthetic PDF vision\/native check failed/, response.body)
assert_match(/Synthetic PDF check passed/, response.body)
assert_match(/Synthetic PDF check failed/, response.body)
assert_match(/Vision\/native failure reason/, response.body)
assert_match(/unexpected response/, response.body)
assert_no_match(/sk-secret-openai/, response.body)
end
test "AI status does not probe PDF processing when it is explicitly disabled" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
AiHealth::Probe.any_instance.expects(:pdf_text_extraction).never
AiHealth::Probe.any_instance.expects(:pdf_vision_processing).never
with_ai_environment(
"OPENAI_ACCESS_TOKEN" => "sk-secret-openai",
"OPENAI_SUPPORTS_PDF_PROCESSING" => "false"
) do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_match(/Disabled or not supported by the effective provider\/model/, response.body)
assert_no_match(/The synthetic PDF .* check failed/, response.body)
end
test "AI status reports Anthropic with an available pgvector store" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
Setting.stubs(:llm_provider).returns("anthropic")
connection = stub("connection")
connection.stubs(:table_exists?).with(VectorStore::Pgvector::TABLE_NAME).returns(true)
connection.stubs(:extension_enabled?).with("vector").returns(true)
ActiveRecord::Base.stubs(:connection).returns(connection)
VectorStore.expects(:embedding_access_token).returns("runtime-embedding-token")
AiHealth::Probe.any_instance.expects(:embedding).with(
endpoint: "http://ollama:11434/v1",
access_token: "runtime-embedding-token",
model: "mxbai-embed-large",
dimensions: 1024
).returns(probe_result(:passing))
with_ai_environment(
"ANTHROPIC_ACCESS_TOKEN" => "anthropic-secret",
"ANTHROPIC_MODEL" => "claude-sonnet-4-6",
"EMBEDDING_URI_BASE" => "http://ollama:11434/v1",
"EMBEDDING_MODEL" => "mxbai-embed-large",
"EMBEDDING_DIMENSIONS" => "1024"
) do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_match(/Anthropic/, response.body)
assert_match(/pgvector/, response.body)
assert_match(/PostgreSQL vector extension/, response.body)
assert_match(/mxbai-embed-large/, response.body)
assert_match(%r{http://ollama:11434/v1}, response.body)
assert_match(/Live checks passed/, response.body)
assert_no_match(/anthropic-secret/, response.body)
end
test "AI status explains when no vector store is configured" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
with_ai_environment do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_match(/No vector store is configured/, response.body)
assert_match(/Uploaded documents cannot be indexed or searched/, response.body)
end
test "AI status marks Qdrant as scaffolded and redacts its URL" do
sign_in users(:sure_support_staff)
stub_healthy_sidekiq
with_ai_environment(
"VECTOR_STORE_PROVIDER" => "qdrant",
"QDRANT_URL" => credentialed_url(
scheme: "https",
host: "qdrant.example.test",
port: 6333,
user: "admin",
password: "qdrant-secret",
query: "api_key=query-secret"
),
"QDRANT_API_KEY" => "header-secret"
) do
get admin_system_health_url(tab: "ai")
end
assert_response :success
assert_match(/Qdrant support is not implemented yet/, response.body)
assert_match(/Scaffolded/, response.body)
assert_match(%r{https://qdrant\.example\.test:6333}, response.body)
assert_no_match(/qdrant-secret|query-secret|header-secret/, response.body)
end
private
def credentialed_url(scheme:, host:, port:, user:, password:, path: nil, query: nil)
URI::Generic.build(
scheme: scheme,
userinfo: "#{user}:#{password}",
host: host,
port: port,
path: path,
query: query
).to_s
end
def probe_result(status, failure_code: nil, http_status: nil)
AiHealth::Probe::Result.new(
status: status,
checked_at: status.in?([ :passing, :failing ]) ? Time.current : nil,
failure_code: failure_code,
http_status: http_status
)
end
def with_ai_environment(overrides = {}, &block)
ClimateControl.modify(AI_ENVIRONMENT.merge(overrides), &block)
end
def stub_healthy_sidekiq
SidekiqHealth.any_instance.stubs(:healthy?).returns(true)
SidekiqHealth.any_instance.stubs(:processes_count).returns(1)
SidekiqHealth.any_instance.stubs(:last_heartbeat_at).returns(Time.current)
SidekiqHealth.any_instance.stubs(:max_queue_latency).returns(0.0)
SidekiqHealth.any_instance.stubs(:enqueued_count).returns(0)
SidekiqHealth.any_instance.stubs(:retry_count).returns(0)
SidekiqHealth.any_instance.stubs(:failed_count).returns(0)
SidekiqHealth.any_instance.stubs(:processed_count).returns(42)
SidekiqHealth.any_instance.stubs(:queue_breakdown).returns([ [ "default", 0, 0.0 ] ])
end
end