A document the model read correctly could still come back as "No reservations
could be extracted". Every type mapper can answer null when the node carries no
reservationFor, or one with no name, and both call sites pushed the item only
if there was one — with no else. The unknown-@type branch next to them does
warn, so a node TREK did not recognise was reported while a node it recognised
but could not map vanished: no item, no warning, nothing in the log. Because
the file did produce a node, the "no reservations found" fallback was skipped
too, and the preview arrived empty and silent. Every recognised type could
disappear this way, on both the kitinerary and the AI path.
Each of them now warns, naming the file, the index, the type and the reason.
Three smaller silences went with it. The debug line labelled "Raw LLM Response"
printed the already-parsed array, so its inspect formatting — single quotes,
bare keys — read as a provider emitting broken JSON and sent this report at the
JSON parser, which has handled that shape since 4.0.0. A single reservation
answered on its own, with no array and no wrapper, was dropped by the list
unwrapper although the schema downstream would have taken it. And a response
nobody could read returned an empty list, which is the same answer as a voucher
with no booking in it; it now raises, on the Anthropic client too, where a
reply cut off at the token limit produced exactly this report's symptom. The
plugin AI surface keeps its old empty answer, since a value is what a plugin
contracts for.
The retry ladder gained a rung that sends no response_format at all, for
providers that support neither json_schema nor json_object.
Closes#2375
With the Anthropic provider, an import sometimes reported that no
reservations were found even though the document held one and the model had
extracted it. Nothing was logged, because from the code's point of view
nothing failed: the value simply was not an array, and the check that only
accepted arrays returned an empty list. What reached the user was
indistinguishable from a document with no booking in it.
The tool declares `reservations` as an array and the call forces the tool, so
it usually is one. Sometimes the model serialises the same content as a
JSON-encoded string instead — either a stringified array, or stringified and
wrapped in `{ reservations: ... }` again. Because that is a serialisation
choice made per call, the same file imported fine on one attempt and came
back empty on the next.
Normalising is now one function next to the lenient JSON parser, and the
OpenAI-compatible client uses it too — it had grown its own version of the
object-unwrap half, and two copies of this rule would drift.
It unwraps a string once and no further. A value that is still not a list
after that is not a list, and reaching deeper would start turning prose into
bookings.
Reported with the exact file and lines, and with both observed shapes.
Closes#1968
Gemini reached through the OpenAI-compatible endpoint returns
JavaScript-object-literal output — single-quoted strings, unquoted keys,
trailing commas — which strict JSON.parse rejects. The reservation list
then came back empty and the booking-import UI showed nothing (#1638).
Add a shared parseLenientJson() that tries strict JSON first and falls
back to JSON5 (which accepts exactly that relaxed superset), and route
both the OpenAI-compatible and Ollama-native parse paths through it,
replacing the two identical strict parsers.
Maurice