Files
sure/test/models/provider/anthropic/message_formatter_test.rb
Guillem Arias 5ab9bb33d6 fix(ai): sanitize Langfuse warn logs, normalize tool_use.input, dedup history fetch
Addresses three open CodeRabbit findings on PR #1983.

- Provider::Anthropic Langfuse rescue branches no longer include
  `e.full_message` in `Rails.logger.warn`. `full_message` bundles the
  backtrace + cause chain and on some SDK error types includes the
  serialized request/response payload (prompt, model output). Logs
  now report `#{e.class}: #{e.message}` only. Three sites:
  create_langfuse_trace, log_langfuse_generation, upsert_langfuse_trace.
  Note: Provider::Openai has the same pattern (copy-pasted source) —
  harmonization deferred to a follow-up cleanup PR; this commit fixes
  only the Anthropic provider to keep PR scope tight.

- MessageFormatter#parse_arguments now coerces any non-Hash parsed
  result to `{}`. Anthropic's Messages API requires `tool_use.input`
  to be a JSON object (map); a stored ToolCall::Function record whose
  arguments parse to a scalar, bool, or array (corrupt row, legacy
  data, cross-provider bleed) would otherwise produce a payload the
  API rejects. Normal flow stores Hash arguments end-to-end so the
  fix is defensive — adds 2 tests covering scalar/array JSON strings
  and non-String non-Hash inputs.

- Assistant::Responder dedups the chat-history fetch. The previous
  layout fired two near-identical `chat.messages.where(...).includes(
  :tool_calls).ordered` queries per LLM turn (one for the OpenAI-shape
  payload, one for the raw-records kwarg). A new memoized
  `complete_chat_messages` fetches once; `chat_message_records` filters
  out the current message via `Array#reject`, `openai_messages_payload`
  iterates the cached array unchanged. One SQL query per turn instead
  of two. Memoization scope = single Responder instance (per LLM call),
  so cache invalidation is not a concern.

All 4370 tests pass (1 pre-existing libvips env error unrelated).
Rubocop + brakeman clean.
2026-05-27 10:42:12 +02:00

238 lines
8.7 KiB
Ruby

require "test_helper"
require "ostruct"
class Provider::Anthropic::MessageFormatterTest < ActiveSupport::TestCase
test "builds a single user turn from prompt alone" do
formatter = Provider::Anthropic::MessageFormatter.new(prompt: "hi")
messages = formatter.build
assert_equal 1, messages.size
assert_equal({ role: "user", content: "hi" }, messages.first)
end
test "skips empty content from history" do
history = [ stub_user_message("") ]
messages = Provider::Anthropic::MessageFormatter.new(prompt: "next", conversation_history: history).build
assert_equal [ { role: "user", content: "next" } ], messages
end
test "renders text-only assistant history with no tool calls" do
history = [
stub_user_message("first question"),
stub_assistant_message("first answer")
]
messages = Provider::Anthropic::MessageFormatter.new(prompt: "second question", conversation_history: history).build
assert_equal({ role: "user", content: "first question" }, messages[0])
assert_equal "assistant", messages[1][:role]
assert_equal [ { type: "text", text: "first answer" } ], messages[1][:content]
assert_equal({ role: "user", content: "second question" }, messages[2])
end
test "renders assistant tool_call history with paired tool_result turn" do
tool_call = stub_tool_call(
id: "toolu_1",
name: "get_net_worth",
arguments: { "currency" => "USD" },
result: { "amount" => 12345, "currency" => "USD" }
)
assistant = stub_assistant_message("Your net worth is $12,345.", tool_calls: [ tool_call ])
history = [ stub_user_message("net worth?"), assistant ]
messages = Provider::Anthropic::MessageFormatter.new(prompt: "anything else?", conversation_history: history).build
assert_equal({ role: "user", content: "net worth?" }, messages[0])
assert_equal "assistant", messages[1][:role]
assert_equal "tool_use", messages[1][:content].first[:type]
assert_equal "toolu_1", messages[1][:content].first[:id]
assert_equal "get_net_worth", messages[1][:content].first[:name]
assert_equal({ "currency" => "USD" }, messages[1][:content].first[:input])
assert_equal "text", messages[1][:content].last[:type]
assert_equal "user", messages[2][:role]
assert_equal "tool_result", messages[2][:content].first[:type]
assert_equal "toolu_1", messages[2][:content].first[:tool_use_id]
assert_equal({ "amount" => 12345, "currency" => "USD" }.to_json, messages[2][:content].first[:content])
assert_equal({ role: "user", content: "anything else?" }, messages[3])
end
test "renders in-flight function_results as assistant tool_use + user tool_result" do
formatter = Provider::Anthropic::MessageFormatter.new(
prompt: "what is my net worth?",
function_results: [ {
call_id: "toolu_42",
name: "get_net_worth",
arguments: { "currency" => "USD" }.to_json,
output: { amount: 99, currency: "USD" }
} ]
)
messages = formatter.build
assert_equal({ role: "user", content: "what is my net worth?" }, messages[0])
assert_equal "assistant", messages[1][:role]
assert_equal "tool_use", messages[1][:content].first[:type]
assert_equal "toolu_42", messages[1][:content].first[:id]
assert_equal({ "currency" => "USD" }, messages[1][:content].first[:input])
assert_equal "user", messages[2][:role]
assert_equal "tool_result", messages[2][:content].first[:type]
assert_equal "toolu_42", messages[2][:content].first[:tool_use_id]
assert_includes messages[2][:content].first[:content], "99"
end
# Confirms the round-trip flagged in PR #1983 review: an Anthropic tool_use
# block returned by the model → ChatFunctionRequest → ToolCall::Function
# persisted on the AssistantMessage → MessageFormatter rebuild on the next
# turn produces an Anthropic-compatible history where tool_use_id pairs back
# to the original block.
test "ChatParser → ToolCall::Function → MessageFormatter round-trips tool_use_id" do
anthropic_response = OpenStruct.new(
id: "msg_abc",
model: "claude-sonnet-4-6",
content: [
OpenStruct.new(type: :tool_use, id: "toolu_round_trip", name: "get_net_worth", input: { "currency" => "USD" })
]
)
parsed = Provider::Anthropic::ChatParser.new(anthropic_response).parsed
function_request = parsed.function_requests.first
persisted_tool_call = ToolCall::Function.from_function_request(
function_request,
{ "amount" => 12345, "currency" => "USD" }
)
assistant = stub_assistant_message("Your net worth is $12,345.", tool_calls: [ persisted_tool_call ])
history = [ stub_user_message("net worth?"), assistant ]
rebuilt = Provider::Anthropic::MessageFormatter.new(prompt: "follow-up", conversation_history: history).build
tool_use_block = rebuilt[1][:content].find { |b| b[:type] == "tool_use" }
tool_result_block = rebuilt[2][:content].first
assert_equal "toolu_round_trip", tool_use_block[:id]
assert_equal "toolu_round_trip", tool_result_block[:tool_use_id]
assert_equal({ "currency" => "USD" }, tool_use_block[:input])
assert_equal({ "amount" => 12345, "currency" => "USD" }.to_json, tool_result_block[:content])
end
test "renders multi-tool assistant turn with all pairings preserved" do
tool_a = stub_tool_call(
id: "toolu_a",
name: "get_accounts",
arguments: {},
result: [ { "id" => 1, "name" => "Checking" } ]
)
tool_b = stub_tool_call(
id: "toolu_b",
name: "get_holdings",
arguments: {},
result: [ { "ticker" => "VTI", "qty" => 10 } ]
)
assistant = stub_assistant_message("Looked up your accounts and holdings.", tool_calls: [ tool_a, tool_b ])
messages = Provider::Anthropic::MessageFormatter.new(
prompt: "follow-up",
conversation_history: [ stub_user_message("accounts and holdings?"), assistant ]
).build
tool_uses = messages[1][:content].select { |b| b[:type] == "tool_use" }
tool_results = messages[2][:content]
assert_equal 2, tool_uses.size
assert_equal 2, tool_results.size
assert_equal [ "toolu_a", "toolu_b" ], tool_uses.map { |b| b[:id] }
assert_equal [ "toolu_a", "toolu_b" ], tool_results.map { |b| b[:tool_use_id] }
# Anthropic requires the user turn to follow the assistant turn that used tools
assert_equal "assistant", messages[1][:role]
assert_equal "user", messages[2][:role]
end
test "parses string arguments and nil outputs gracefully" do
formatter = Provider::Anthropic::MessageFormatter.new(
prompt: "go",
function_results: [ {
call_id: "toolu_x",
name: "noop",
arguments: "",
output: nil
} ]
)
messages = formatter.build
assert_equal({}, messages[1][:content].first[:input])
assert_equal "", messages[2][:content].first[:content]
end
# Anthropic's tool_use.input MUST be a JSON object (map). If a stored
# ToolCall::Function record carries arguments that parse to a scalar or
# array (corrupt row, legacy data, OpenAI cross-bleed), the formatter
# must coerce them to `{}` so we don't ship an invalid payload.
test "coerces non-Hash parsed arguments to empty Hash" do
[ '"hello"', "123", "true", "[1,2,3]" ].each do |non_object_json|
formatter = Provider::Anthropic::MessageFormatter.new(
prompt: "go",
function_results: [ {
call_id: "toolu_x",
name: "noop",
arguments: non_object_json,
output: nil
} ]
)
messages = formatter.build
assert_equal({}, messages[1][:content].first[:input],
"expected empty Hash for arguments=#{non_object_json.inspect}")
end
end
test "coerces non-Hash non-String arguments to empty Hash" do
formatter = Provider::Anthropic::MessageFormatter.new(
prompt: "go",
function_results: [ {
call_id: "toolu_x",
name: "noop",
arguments: [ 1, 2, 3 ],
output: nil
} ]
)
messages = formatter.build
assert_equal({}, messages[1][:content].first[:input])
end
private
def stub_user_message(content)
msg = UserMessage.new(content: content, ai_model: "claude-sonnet-4-6")
msg.id = SecureRandom.uuid
msg
end
def stub_assistant_message(content, tool_calls: [])
msg = AssistantMessage.new(content: content, ai_model: "claude-sonnet-4-6")
msg.id = SecureRandom.uuid
msg.stubs(:tool_calls).returns(tool_calls)
msg
end
def stub_tool_call(id:, name:, arguments:, result:)
tc = ToolCall::Function.new(
function_name: name,
function_arguments: arguments,
function_result: result
)
tc.stubs(:provider_call_id).returns(id)
tc.stubs(:provider_id).returns(id)
tc
end
end