Enum WebStack 

pub enum WebStack {
    Spine,
    OpenAiApi,
    AnthropicApi,
    Mcp,
    Grpc,
    HttpJson,
    GraphQl,
}

Web stacks / wire protocols with curated agentic profiles.

Variants

Spine

SPINE (nervosys/SPINE) — the agentic-first web stack. Native StreamStart/Token/End frames with text / bytes / encoded-latent / tool-call data variants; CapabilityQuery/CapabilityAdvertisement with JSON Schema and optional semantic embeddings; TraceContext (W3C traceparent) attached inline; OpenAI-compatible SSE bridge in the gateway; bearer auth secure by default as of v1.3.0; optional FIPS 140-3 build via aws-lc-rs.

OpenAiApi

OpenAI API — HTTP + JSON with SSE chat.completion.chunk streaming; tools parameter for function calling; bearer auth + TLS.

AnthropicApi

Anthropic API — HTTP + JSON with SSE message_start/delta/stop streaming; tools parameter for tool use; bearer auth + TLS.

Mcp

Model Context Protocol — JSON-RPC over stdio or SSE; tools/list, resources/list, prompts/list introspection RPCs are the surface.

Grpc

gRPC — protobuf over HTTP/2 with first-class server / client / bidi streaming, service reflection, mTLS, interceptors for auth/tracing.

HttpJson

Plain HTTP + JSON (REST-shaped). The generic baseline an agent has to fall back to when nothing more specific exists.

GraphQl

GraphQL — query language with introspection built into the protocol; subscriptions for streaming.

Implementations

impl WebStack

pub fn all() -> [WebStack; 7]

All profiled web stacks, in fixed (deterministic) order.

pub fn name(self) -> &'static str

Canonical lowercase name.

Examples found in repository

examples/web_benchmark.rs (line 25)

fn main() {
    println!("agentic-eval — web stacks / wire protocols for agentic AI use");
    println!("axes: streaming, tool-discoverability, encoding, interop, security\n");

    // ── Ranked benchmark (best-first by composite agentic fitness) ───────────
    println!(
        "{:<15} {:>7}   {:>9} {:>5} {:>8} {:>7} {:>8}",
        "stack", "fitness", "streaming", "tools", "encoding", "interop", "security"
    );
    for p in rank_web_stacks() {
        println!(
            "{:<15} {:>7.2}   {:>9.2} {:>5.2} {:>8.2} {:>7.2} {:>8.2}",
            p.stack.name(),
            p.fitness(),
            p.streaming,
            p.tool_discoverability,
            p.encoding_efficiency,
            p.interop,
            p.security_primitives,
        );
    }

    // ── Head-to-head: SPINE vs the OpenAI API (the dominant baseline) ───────
    println!("\nhead-to-head (positive = SPINE fits agentic use better):");
    print!("{}", compare_web_stacks(WebStack::Spine, WebStack::OpenAiApi));

    // ── Evidence behind SPINE's profile ─────────────────────────────────────
    println!("\nwhy SPINE scores where it does:");
    for e in &profile(WebStack::Spine).evidence {
        println!("  - {e}");
    }

    println!(
        "\nReading: SPINE now leads the composite (0.90), edging gRPC (0.83).\n\
         It was always strong on the agent-native axes it was designed for\n\
         (LLM StreamStart/Token/End frames with multiplex-aware StreamCancel +\n\
         mid-stream usage, a CapabilityQuery handshake, inline W3C TraceContext).\n\
         v1.4.0's CBOR wire format plus v1.5.0's byte-string tensor payloads\n\
         bring encoding to parity with protobuf (0.95; 89% smaller embedding\n\
         frames), and per-message Ed25519 signed frames give message-level\n\
         non-repudiation beyond channel mTLS (security 0.95). The inflection is\n\
         three deployable bridges: a runnable MCP stdio server (v1.6.0), the\n\
         OpenAI-compatible gateway, and a production-grade gRPC AgentService\n\
         (v1.8.0, reflection-enabled + real-model-backed in v1.9.0) — reachable\n\
         from the three dominant agent ecosystems with standard client stubs,\n\
         lifting interop 0.15 -> 0.67. Honest\n\
         caveat: interop is still SPINE's weakest axis — the bridges map the\n\
         agentic surface (not SPINE's native binary frames) and SPINE's own\n\
         protocol has ~zero native install base."
    );
}

pub fn from_name(name: &str) -> Option<WebStack>

Parse a (case-insensitive) name; accepts common aliases (openai, claude, model-context-protocol, rest, graphql-spec, …).

Trait Implementations

impl Clone for WebStack

fn clone(&self) -> WebStack

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for WebStack

impl Debug for WebStack

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for WebStack

impl Hash for WebStack

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for WebStack

fn eq(&self, other: &WebStack) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for WebStack

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for WebStack