recommend_runtimeChoisit un chemin runtime à partir d'un OS, d'une VRAM et d'une RAM déclarés. Ce n'est pas une installation, pas un scan et pas un benchmark.
| Name | Required | Description | Default |
|---|---|---|---|
| usage | No | polyvalent | |
| ram_gb | No | ||
| os_name | No | ||
| vram_gb | No | ||
| unified_memory | No |
| Name | Required | Description | Default |
|---|---|---|---|
| runtime | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and destructiveHint=false, so the description's main behavioral contribution is the negative clarification (not a scan/benchmark). The description does not disclose any additional behavioral traits such as expected output format, error handling, or whether the tool requires internet access. With annotations covering the safety profile, the description adds only marginal value.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, no filler, front-loaded with the purpose. Every sentence adds value. It is appropriately sized for the tool's complexity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 5 parameters, an output schema, and multiple siblings, the description is brief but covers the core purpose. However, it fails to explain the 'usage' and 'unified_memory' parameters, offers no examples, and does not clarify how the tool relates to siblings. It is adequate but has clear gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0% for 5 parameters, so the description must compensate. It mentions three of the five parameters (OS, VRAM, RAM) but omits 'usage' and 'unified_memory', and does not explain the enum values for 'usage'. The description adds some meaning beyond the schema, but incomplete coverage leaves the agent guessing about the missing parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('chooses a runtime path') and the key inputs (OS, VRAM, RAM), and it explicitly excludes other actions (installation, scan, benchmark). The title specifies the runtimes (Ollama, LM Studio, llama.cpp). However, it does not differentiate from sibling tools like check_pc_for_local_ai or list_models_for_budget, leaving some ambiguity about when to use this tool over those.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description only provides negative guidance: what the tool is NOT (installation, scan, benchmark). It offers no positive guidance on when to use this tool vs. alternatives, no prerequisites, and no context about the recommended use case. This is insufficient for an agent to decide when to invoke this tool.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
Each tool has a distinct purpose (analyze, check, explain, list, etc.) and the detailed descriptions make boundaries clear. However, the difference between 'reading a shared report' (analyze_shared_report) and 'reading a shared report to list installed models' (list_installed_models_from_report) or 'reading a shared report to list benchmarks' (list_benchmark_proofs_from_report) could cause an agent to pick the wrong one. There is also some overlap between these list operations that read a report versus simply displaying a pre-generated cockpit (render_machine_cockpit).
The majority of tools follow a consistent verb_noun pattern (e.g., analyze_shared_report, explain_bottleneck, list_installed_models_from_report, recommend_runtime). Minor inconsistency exists with the use of 'geo_audit' vs 'geo_kit' vs 'ratings' and the verb tense in 'list_benchmark_proofs_from_report' and 'list_first_party_measurements' departs from a simple pattern.
15 tools is at the high end of the ideal range (3-15), but each tool appears justified given the comprehensive scope of local AI assistance (hardware checking, model lookup, benchmarking, reporting, educational/reference tools). Slightly over-stuffed but still manageable for an agent.
The surface covers a complete workflow: check hardware, lookup models, benchmark, generate cockpit, explain bottlenecks, simulate upgrades, and reference documentation. The missing piece is the lack of 'update' or 'delete' operations, but this is by design, as the entire workflow is read-only. The set seems like a complete view of all possible read-only interactions with the domain.