Skip to main content
Glama

justicelibre

by io.github.Dahliyaal

search_admin

Read-only

Recherche pondérée par pertinence BM25 sur la jurisprudence administrative complète (Conseil d'État + Tribunal des conflits + 9 CAA + 40 TA — pour le TC, filtrer avec juridiction="conflits").

Source : bulk JADE DILA (~550 k décisions full text). Contrairement aux
outils `search_admin_recent*` qui trient par date, celui-ci classe par
pertinence sémantique des mots-clés. Indispensable pour trouver LES
bonnes décisions sur un sujet sans dépendre de l'ancienneté.

⚠️ **Si tu cherches par numéro de requête (7 chiffres ex: 2200433)**,
utilise plutôt `get_admin_decision(numero, juridiction=...)` qui fait
un lookup SQL exact. La recherche FTS5 d'un numéro court ne le trouve
que dans les décisions qui le **citent** dans leur texte (ex: décision
de cassation), pas la décision identifiée par ce numéro.

Args:
    query: mots-clés (opérateurs FTS5 : AND/OR/NOT, "phrase exacte", mot*)
    juridiction: ⚠️ **ce n'est PAS un filtre** — malgré son nom. Le
        fragment est ajouté à la requête plein texte en `AND`, donc il
        matche aussi les décisions qui se contentent de **citer** cette
        juridiction. Mesuré le 23 août 2026 :
        `juridiction="Tribunal Administratif de Melun"` renvoie 1 506
        résultats dont AUCUN des cinq premiers n'émane de Melun (des CAA
        de Paris et un CE, qui statuaient sur des appels venus de Melun).
        Le `total` n'est donc pas la volumétrie de la juridiction.
        Utilise-le pour ORIENTER la recherche, jamais pour t'assurer de
        l'origine d'une décision : vérifie le champ `juridiction` de
        chaque résultat. Pour un tri réellement par juridiction, passer
        par `search_admin_recent(juridiction_code=…)` ou
        `get_admin_decision(numero, juridiction=…)`, qui, eux, contrôlent
        l'ordre de juridiction.
    sort: "relevance" (défaut, BM25) ou "date_desc" / "date_asc"
    date_min: limite inférieure ISO YYYY-MM-DD (optionnel)
    date_max: limite supérieure ISO YYYY-MM-DD (optionnel)
    limit: nombre de résultats (défaut 20, max 50)
    offset: pagination

Returns:
    {"total", "returned", "decisions": [...]} avec extracts BM25.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
sortNorelevance
limitNo
queryYes
offsetNo
date_maxNo
date_minNo
juridictionNo

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault

No arguments

TDQS

A4.9/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description adds substantial non-obvious behavior far beyond the readOnlyHint annotation: the juridiction parameter is disclosed as NOT a filter, with a concrete measured example explaining it matches citations, and the meaning of `total` is clarified. This is highly useful behavioral context that cannot be derived from the schema.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is long (~300 words) and intentionally repeats the juridique warning across multiple sections, which is justified given the risk of misuse. The core purpose is front-loaded in the first sentence, and each paragraph serves one distinct purpose, though the measured statistics example could be slightly more trimmed without loss of essential information.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity and the presence of an output schema, the description is essentially complete: parameter semantics, date filtering, sort keys, limits, alternatives by scenario, jurisdiction caveats, and expected return shape all are present. Any missing details would be redundant given the output schema.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, so the description fully compensates: it documents all 7 parameters, the FTS5 operators (AND/OR/NOT, exact phrase, mot* — though encoding of 'mot*' is an approximation), the default sort values, the ISO date format, default and max limit (20/50), offset pagination, and the highly misleading juridiction argument with detailed warning and expected behavior.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb + resource: 'Recherche pondérée par pertinence BM25 sur la jurisprudence administrative complète' with the scope enumerated (CE, TC, CAA, TA, ~550k décisions). It explicitly positions itself against search_admin_recent* siblings and get_admin_decision, so an agent can distinguish it from the 28 siblings without opening schemas.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives explicit when-to-use ('Indispensable pour trouver LES bonnes décisions sur un sujet sans dépendre de l'ancienneté'), explicit when-not-to-use for request-number lookup ('utilise plutôt get_admin_decision(numero, juridiction=...)'), and names alternatives for true jurisdiction filtering (search_admin_recent, get_admin_decision). Nothing is left to inference.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All justicelibre toolsInspect
Try in Browser
RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A4.1/5.0
Disambiguation4/5

The tools are mostly distinct by legal source (admin, CC, CEDH, CJUE, etc.) and by action (search vs. retrieve). However, pairs like search_admin vs. search_conseil_etat both target Council of State jurisprudence, and get_decision_text could be mistaken for a generic decision retriever, creating some potential confusion despite detailed descriptions.

Naming Consistency5/5

All tool names follow a consistent lowercase snake_case pattern with action-first naming (get_, search_, list_, build_, resolve_, about_). The get_decision_* and search_* prefixes clearly separate retrieval from search, and modifiers like _libre and _recent are applied systematically.

Tool Count2/5

With 31 tools, the server is heavy, exceeding the 25+ threshold for 'too many.' While the breadth of French legal sources justifies many search/retrieval tools, the three recent admin variants and the redundant PISTE vs. libre duplicates suggest the set could be consolidated. The count feels disproportionate to the core purpose of legal research.

Completeness4/5

The tool set covers the full read lifecycle for legal research: orientation (about), search across all major sources, full-text retrieval for each source, law article lookup with historical versions, and URL building. Minor gaps exist, such as lacking a structured code browse or a direct way to enumerate all articles in a section, but these are workarounds via search_legi and get_law_article.

Resources