Archival Summary

🤖 AI Drafted (Not reviewed)

Compose nine-section archival summary for a folder (leaf node of the Catalogue workflow)

Tool id	catalogue
Category	llm
Uses a language model	yes
Needs a generative model	no
Runs over many items	no
Item handling	batch
Structured output	yes
Human-verified	not yet

What it reads

Port	Type	Required	What it is
Text (text)	text	no	Aggregated TRANSCRIPTION text — wire from aggregate.text (per-file transcriptions joined), NOT from the merge of cleanup outputs. The narrative LLM reads this as source material; pointing it at cleaned entity lists makes it summarise the cleanup output instead of the documents.
Data (data)	any	yes	Dependency-only input. Its value is ignored; it exists to sequence catalogue after its upstream — canonical entity cleanup in the full preset, the source node in ‘6 · Catalogue’.
Context (context)	any	no	Previous text/transcription
Metadata (metadata)	json	no	Existing metadata
Documents (documents)	json	no	Document metadata

What it emits

Port	Type	Required	What it is
Text (text)	text	—	Raw text response
Value (value)	any	—	Parsed value
Texts (texts)	array	—	Per-item texts
Values (values)	array	—	Per-item values
Results (results)	json	—	Full results
Records (records)	array	—	Per-document text records [{doc_id, text}, …].
Artifacts (artifacts)	json	—	Artifact IDs

Options

Option	Type	Default	What it does
ambiguous_min_items	integer	25	Minimum child documents before ambiguity HITL can trigger.
choices	array	—	Valid choices. (Not shown in the editor.)
chunk_size_chars	integer	0	Chunk large input text above this character budget (0=auto).
claim_context_caps	object	{}	Per-section caps for the entity-context block fed to the narrative LLM. Keys: people, places, organizations, events, dates, keywords. Each value is the max items to include before the list is truncated with ‘(… +N more)’. Defaults: people=30, places=20, organizations=15, events=15, dates=30, keywords=30. Override only the sections you want to change. Useful on dense folders (100+ contributors) or sparse letters where you want to dial cap up/down. (#865).
hitl_on_ambiguous_grouping	boolean	no	Pause for confirmation when case grouping appears ambiguous.
match_mode	string	prefer	Match mode. One of: prefer, strict, inform.
max_items	integer	10	List max items.
max_tokens	integer	8192	Max response.
max_words	integer	50	Word limit.
metadata_field	string	—	Save to field.
model_name	string	—	Model name.
output_format	string	text	Response format. One of: text, boolean, choice, number, words, list, json.
output_language	string	auto	Output language. ‘auto’ detects from the source text (English / Spanish today); explicit names like ‘English’ or ‘Spanish’ pin the language regardless of input.
prompt	string	—	Custom prompt.
provider_name	string	—	LLM provider. One of: openai, anthropic, google, ollama, lmstudio, groq, together, deepseek, mistral, openrouter, dashscope, xai, perplexity, fireworks, deepl.
quality_gate	boolean	yes	Stop the run if output is unreadable.
reference_values	object	—	Known values to match. (Not shown in the editor.)
save_to_db	boolean	yes	Save to library.
save_to_file	boolean	no	Export to file.
temperature	number	0.7	Creativity.
thinking_mode	string	off	Chain-of-thought reasoning depth. One of: off, short, medium, long.

The prompt it sends

This is what the tool asks a model, with every option left at its default. Changing the options above changes this text.

You are an expert archivist. Write a catalogue entry in English
for what these documents CONTAIN. One paragraph, plain prose. NO title,
NO heading, NO label like "Catalogue Entry" or "Summary:" — start the
entry directly with the document type. NO bold markers (**…**), NO
Markdown headers (#), NO bullets, NO JSON.

Open with the document type in the source's own vocabulary (a deed,
lawsuit, letter, report, chapter, photograph, etc.). Length matches
the source: 30-100 words for a short formulaic record (parties +
object + price + terms); 200-450 words for a long file (subject and
actors first, then concrete dates, places, sums, occupations, claims,
outcomes).

Use evidentiary verbs: contains, records, states, alleges, describes,
names, signs. Preserve concrete details verbatim — names, dates, sums,
places, terms, injuries, sentences. Preserve any [ilegible] /
[uncertain] markers and accents verbatim; do not smooth them away.
Frame any racial, caste, or status label as the source's:
"(described as …)" / "(caracterizado como …)".

Do not invent names, dates, places, or facts. Do not interpret. Do
not add atmosphere, mood, theme, or significance. Plain working prose.