@tricoteuses/tisseuse 0.13.8 → 0.13.10

package/README.md

@@ -96,6 +96,17 @@ npm run index:postgres:assemblee -- --all
 
 La première commande consomme la file des documents à acquérir, simplifier et enrichir. La seconde indexe dans `tisseuse.assemblee_text_search` les articles et les exposés des motifs comme résultats distincts. Les déclencheurs sur `assemblee.documents`, le hash du contenu et les files PostgreSQL permettent les mises à jour ciblées ; les options `--all` demandent une réconciliation complète.
 
+Pour acquérir et indexer immédiatement quelques documents indépendamment de leur position dans les files durables :
+
+```sh
+npm run sync:postgres:assemblee-content -- \
+  --document-uids PRJLANR5L17BTC2765,PRJLANR5L17BTC3067
+npm run index:postgres:assemblee -- \
+  --document-uids PRJLANR5L17BTC2765,PRJLANR5L17BTC3067
+```
+
+Le mode ciblé refuse les UID absents de `assemblee.documents` et ne peut pas être combiné avec `--all`.
+
 Exploratrice et Moulineuse ne téléchargent jamais un document pendant une lecture. Lorsqu’un document source existe mais que son contenu n’est pas encore disponible, son état est conservé en base et l’interface doit l’indiquer au lieu de servir une copie locale périmée.
 
 ### Première itération de recherche plein texte Sénat
@@ -133,46 +144,34 @@ Les documents sans URL, externes au Sénat, sans Akoma Ntoso ou en échec de par
 
 ## Legal reference resolver
 
-Tisseuse resolves legal references through a `LegiResolver`. The default
-resolver is `canutes`, which uses the configured Canutes database and a generated
-`text_titles_infos.json` file.
+Tisseuse resolves legal references through the Canutes-backed `LegiResolver`.
+`LegiResolver` is exported as an abstract class from both
+`@tricoteuses/tisseuse` and `@tricoteuses/tisseuse/server`, so applications can
+provide another storage-backed implementation. At process startup, the Canutes
+resolver reads the title metadata it needs directly from
+`legifrance.texte_version` and builds the parser indexes in memory. Concurrent
+callers share the same preload operation, and the resulting indexes remain fixed
+for the lifetime of the process.
 
-There are two supported resolver paths for tests and scripts:
+No generated title index belongs in the source tree. Tests that resolve the full
+Légifrance title corpus need access to the configured Canutes database and call
+the same preload operation as production code.
 
-- `f_legi`: uses the flat f_legi database and does not need a generated
-  `text_titles_infos.json` file.
-- `canutes`: uses the Canutes database and requires generating
-  `src/lib/text_parsers/text_titles_infos.json` before running tests that parse
-  or resolve French text titles.
+### `f_legi` server adapter
 
-Use `f_legi` explicitly when you want to run against the flat f_legi database:
+Applications backed by a `f_legi` database can import the optional server-only
+adapter from the same package:
 
-```sh
-TISSEUSE_LEGI_RESOLVER=f_legi npm test
-npx tsx src/scripts/parse_assemblee_bill_raw.ts bill.html.raw --resolver f_legi
-```
+```ts
+import { FlegiResolver } from "@tricoteuses/tisseuse/server/f-legi"
 
-For the default Canutes resolver, generate the text title index first, then run
-the tests:
-
-```sh
-npx tsx src/scripts/extract_texts_titles_infos.ts
-npm test
+const resolver = new FlegiResolver(sql)
 ```
 
-This creates `src/lib/text_parsers/text_titles_infos.json`. It is generated from
-the Canutes database, is listed in `.gitignore`, and must never be committed.
-
-If you need to keep the generated file outside the working tree, pass it through
-`TISSEUSE_TEXT_TITLES_INFOS_PATH` or script options instead:
-
-```sh
-TISSEUSE_TEXT_TITLES_INFOS_PATH=/path/to/text_titles_infos.json npm test
-npx tsx src/scripts/parse_akoma_ntoso_bill_raw.ts bill.akn.xml.raw --text-titles-infos /path/to/text_titles_infos.json
-```
+The caller owns the PostgreSQL client and its lifecycle. Importing the adapter
+does not open a connection, read environment variables, or preload title data.
+The generic Tisseuse entry points do not load this adapter.
 
-If no path is supplied, the Canutes resolver first tries
-`src/lib/text_parsers/text_titles_infos.json` from the current working tree, then
-the same generated file in the local tisseuse package source tree. That file is
-generated data and is not loaded through Vitest global setup; tests that need it
-preload it explicitly through the selected resolver.
+External adapters can extend the public `LegiResolver` class. The public package
+also exports the resolver option types and `buildAndInstallTextTitlesInfos`, so
+an adapter does not need deep imports into Tisseuse.