server: share JSON values through a named module

build.zig

@@ -11,6 +11,7 @@ const grammar_build = @import("zig/build/grammar.zig");
 const counters_build = @import("zig/build/metal_counters.zig");
 const exl3_build = @import("zig/build/exl3.zig");
 const exlr8_build = @import("zig/build/exlr8.zig");
+const server_build = @import("zig/build/server.zig");
 const golden_build = @import("zig/build/golden.zig");
 
 comptime {
@@ -204,6 +205,7 @@ fn nativeServer(b: *std.Build, target: std.Build.ResolvedTarget, optimize: std.b
         .imports = &.{ .{ .name = "engine_api", .module = api }, .{ .name = "tokenizer", .module = tokenizer }, .{ .name = "template", .module = template }, .{ .name = "native_engines", .module = engines }, .{ .name = "checkpoint_cli", .module = b.createModule(.{ .root_source_file = b.path("zig/src/cli/cli.zig"), .target = target, .optimize = .ReleaseSafe, .link_libc = true, .imports = &.{ .{ .name = "native_engines", .module = engines }, exlr8_build.cliImport(b, target, .ReleaseSafe, engine) } }) } },
     }) });
     exe.root_module.addOptions("build_options", build_options);
+    _ = server_build.addTo(b, exe.root_module, target, .ReleaseSafe);
     if (test_step == null) {
         exe.root_module.strip = true;
         return exe;
@@ -221,10 +223,13 @@ fn nativeServer(b: *std.Build, target: std.Build.ResolvedTarget, optimize: std.b
         .link_libc = true,
         .imports = &.{ .{ .name = "engine_api", .module = api }, .{ .name = "tokenizer", .module = tokenizer }, .{ .name = "template", .module = template } },
     });
+    const server_json = server_build.addTo(b, server_tests, target, .Debug);
     test_step.?.dependOn(&b.addRunArtifact(b.addTest(.{ .root_module = server_tests })).step);
+    test_step.?.dependOn(&b.addRunArtifact(b.addTest(.{ .root_module = server_json })).step);
     @import("zig/build/cortheon_fuzz.zig").add(b, target, api, tokenizer, template);
     // the parse half's tests, as their own test binary (zig/src/server/tool_parse_test.zig)
     const tool_parse_tests = b.addTest(.{ .root_module = b.createModule(.{ .root_source_file = b.path("zig/src/server/tool_parse_test.zig"), .target = target, .optimize = .Debug, .link_libc = true, .imports = &.{ .{ .name = "engine_api", .module = api }, .{ .name = "tokenizer", .module = tokenizer }, .{ .name = "template", .module = template } } }) });
+    _ = server_build.addTo(b, tool_parse_tests.root_module, target, .Debug);
     test_step.?.dependOn(&b.addRunArtifact(tool_parse_tests).step);
     // the chat-template engine's tests (zig/src/core/template/tests.zig), which no other test binary reaches
     test_step.?.dependOn(&b.addRunArtifact(b.addTest(.{ .root_module = b.createModule(.{ .root_source_file = b.path("zig/src/core/template/tests.zig"), .target = target, .optimize = .Debug, .link_libc = true }) })).step);

zig/build/cortheon_fuzz.zig

@@ -20,6 +20,7 @@ pub fn add(b: *std.Build, target: std.Build.ResolvedTarget, api: *Module, tokeni
         }),
         .filters = &.{"cortheon.fuzz."},
     });
+    _ = @import("server.zig").addTo(b, harness.root_module, target, optimize);
     const step = b.step("fuzz-cortheon", "Cortheon's page, link, search and filter readers under the fuzz harness");
     step.dependOn(&b.addRunArtifact(harness).step);
 }

zig/build/cuda.zig

@@ -173,6 +173,7 @@ fn nativeServer(b: *std.Build, target: std.Build.ResolvedTarget, optimize: std.b
         .imports = &.{ .{ .name = "engine_api", .module = m.api }, .{ .name = "tokenizer", .module = tokenizer }, .{ .name = "template", .module = template }, .{ .name = "native_engines", .module = m.engines }, .{ .name = "checkpoint_cli", .module = checkpointCli(b, target, .ReleaseSafe, null, m.engines, core) } },
     }) });
     exe.root_module.addOptions("build_options", build_options);
+    _ = @import("server.zig").addTo(b, exe.root_module, target, .ReleaseSafe);
     if (!install_native) {
         exe.root_module.strip = true;
         return exe;
@@ -201,9 +202,14 @@ pub fn hostTests(b: *std.Build, draft_ids: *std.Build.Module, build_options: *st
         .link_libc = true,
         .imports = &.{ .{ .name = "engine_api", .module = native_modules.api }, .{ .name = "tokenizer", .module = mods.tokenizer }, .{ .name = "template", .module = template } },
     });
+    const server_json = @import("server.zig").addTo(b, server_tests, host, .debug);
     const server_test = b.addRunArtifact(b.addTest(.{ .root_module = server_tests }));
-    step.dependOn(&server_test.step);
-    b.step("test-server-cpu", "The HTTP server's unit tests (routes, templates, tool parsing) without a GPU").dependOn(&server_test.step);
+    const json_test = b.addRunArtifact(b.addTest(.{ .root_module = server_json }));
+    const server_step = b.step("test-server-cpu", "The HTTP server's unit tests (routes, templates, tool parsing) without a GPU");
+    for ([_]*std.Build.Step.Run{ server_test, json_test }) |run| {
+        step.dependOn(&run.step);
+        server_step.dependOn(&run.step);
+    }
     // the server goldens and Cortheon's fuzz step here when the target is not macOS, whose Metal build defines them
     if (server_steps) {
         golden_build.step(b, host, server_tests, native_modules.api, mods.tokenizer, template);

zig/build/server.zig

@@ -0,0 +1,14 @@
+//! The server's shared JSON values, available to server and family modules.
+const std = @import("std");
+
+/// The server's JSON values as a module (zig/src/server/json.zig), so a family's serving code shares them.
+pub fn json(b: *std.Build, target: std.Build.ResolvedTarget, optimize: std.builtin.OptimizeMode) *std.Build.Module {
+    return b.createModule(.{ .root_source_file = b.path("zig/src/server/json.zig"), .target = target, .optimize = optimize });
+}
+
+/// Attach a server's shared JSON module; its tests run separately from the server's root.
+pub fn addTo(b: *std.Build, m: *std.Build.Module, target: std.Build.ResolvedTarget, optimize: std.builtin.OptimizeMode) *std.Build.Module {
+    const j = json(b, target, optimize);
+    m.addImport("json", j);
+    return j;
+}

zig/src/server/anthropic.zig

@@ -1,6 +1,6 @@
 //! Anthropic's Messages routes over the chat handler: typed SSE events, tool_use blocks, count_tokens.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const http_body = @import("http_body.zig");
 const fields = @import("fields.zig");

zig/src/server/anthropic_translate.zig

@@ -1,6 +1,6 @@
 //! Anthropic Messages requests as the chat requests that run them (``anthropic_translate.translate``).
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields = @import("fields.zig");
 const pyrepr = @import("pyrepr.zig");

zig/src/server/chat.zig

@@ -1,7 +1,7 @@
 //! One reply, as ``ChatApp.chat`` makes it: render, submit to the engine, stream text, reasoning and calls.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields_mod = @import("fields.zig");
 const reply_text = @import("reply_text.zig");

zig/src/server/chunk_plan.zig

@@ -1,6 +1,6 @@
 //! Where a prompt's prefill chunks start (Python's PrefillPlan and message_markers): prompts that agree up to a start cut alike up to it.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const prompt = @import("prompt.zig");
 const Server = @import("server.zig").Server;

zig/src/server/compact.zig

@@ -1,6 +1,6 @@
 //! Note and reply are ordinary greedy or keyed generations, so the same request compacts and answers the same way.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const chat = @import("chat.zig");
 const stored = @import("compact_mem.zig");
 const server_mod = @import("server.zig");

zig/src/server/compact_mem.zig

@@ -1,6 +1,6 @@
 //! The compaction note file and the hash of the messages a note already covers.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const Server = @import("server.zig").Server;
 const Value = json.Value;
 const Allocator = std.mem.Allocator;

zig/src/server/compact_test.zig

@@ -1,7 +1,7 @@
 //! Host checks for opt-in compaction. The fake engine records each prompt and answers OK.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const model_text = @import("model_text.zig");
 const chat = @import("chat.zig");

zig/src/server/completion_prompt_test.zig

@@ -1,7 +1,7 @@
 //! Token ids in a completion prompt, top-level or nested: ids in the vocabulary are served, other values get a 400.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const model_text = @import("model_text.zig");
 const openai = @import("openai.zig");
 const Server = @import("server.zig").Server;

zig/src/server/cortheon/checks.zig

@@ -1,6 +1,6 @@
 //! Cortheon's checks before a reply ends: changed files tested before "done", researched answers naming sources.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const tools = @import("tools.zig");
 const Value = json.Value;
 

zig/src/server/cortheon/cortheon.zig

@@ -1,6 +1,6 @@
 //! Cortheon in the server (--cortheon): its reasoning program, web research before building, server-run web tools.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const chat = @import("../chat.zig");
 const grammar = @import("../grammar.zig");
 const prompt_mod = @import("../prompt.zig");

zig/src/server/cortheon/cortheon_test.zig

@@ -1,7 +1,7 @@
 //! Cortheon end to end on a scripted engine and a stand-in web: rounds, research, streams, splices and checks.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("../json.zig");
+const json = @import("json");
 const model_text = @import("../model_text.zig");
 const openai = @import("../openai.zig");
 const server_mod = @import("../server.zig");

zig/src/server/cortheon/loop.zig

@@ -1,6 +1,6 @@
 //! Cortheon's reply loop: rounds with the model's server tool calls run between them, checks at the end, one reply.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const errors = @import("../errors.zig");
 const chat = @import("../chat.zig");
 const tool_parse = @import("../tool_parse.zig");

zig/src/server/cortheon/memory.zig

@@ -1,6 +1,6 @@
 //! What a reply's client never saw (server tool rounds, Cortheon notes), kept to splice back in front of it next turn.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const Value = json.Value;
 const Allocator = std.mem.Allocator;
 

zig/src/server/cortheon/research.zig

@@ -1,6 +1,6 @@
 //! Research before building: the model names searches, the server runs them and reads the best pages, as hidden rounds.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const errors = @import("../errors.zig");
 const chat = @import("../chat.zig");
 const ids = @import("../ids.zig");

zig/src/server/cortheon/search.zig

@@ -1,6 +1,6 @@
 //! Web search for Cortheon: DuckDuckGo's keyless HTML results, or Brave's search API when a key is set.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const html = @import("html.zig");
 const Allocator = std.mem.Allocator;
 

zig/src/server/cortheon/tools.zig

@@ -1,6 +1,6 @@
 //! Cortheon's server-run tools: their specs, which calls are theirs, and each call run into the text the model reads.
 const std = @import("std");
-const json = @import("../json.zig");
+const json = @import("json");
 const search = @import("search.zig");
 const Seen = @import("seen.zig").Seen;
 const Value = json.Value;

zig/src/server/decisions.zig

@@ -1,6 +1,6 @@
 //! /v1/decisions checks a body the way 0.6.6 checks it, then scores through the engine hook when a family sets one.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields = @import("fields.zig");
 const pyrepr = @import("pyrepr.zig");

zig/src/server/decisions_test.zig

@@ -1,7 +1,7 @@
 //! Request checks for /v1/decisions, and a fake engine hook that scores one choice.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const openai = @import("openai.zig");
 const model_text = @import("model_text.zig");

zig/src/server/fields.zig

@@ -1,6 +1,6 @@
 //! Request field checks shared by every route: numbers, stops, thinking switches, efforts, probabilities.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const Value = json.Value;
 const Cx = errors.Cx;

zig/src/server/grammar.zig

@@ -1,7 +1,7 @@
 //! Structured-output fields checked as ``engine.grammar.request_spec`` checks them; the engine enforces them.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const pyrepr = @import("pyrepr.zig");
 const tool_specs = @import("tool_specs.zig");

zig/src/server/hf_text.zig

@@ -2,7 +2,7 @@
 const std = @import("std");
 const tokenizer = @import("tokenizer");
 const template = @import("template");
-const json = @import("json.zig");
+const json = @import("json");
 const model_text = @import("model_text.zig");
 const Value = json.Value;
 const Allocator = std.mem.Allocator;

zig/src/server/http_body.zig

@@ -1,6 +1,6 @@
 //! Bounded request bodies, fixed-length or chunked, refused as the Python server refuses them.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const Conn = @import("http_conn.zig").Conn;
 const Allocator = std.mem.Allocator;

zig/src/server/json.zig

@@ -5,6 +5,7 @@ const Allocator = std.mem.Allocator;
 pub const parse = @import("json_parse.zig").parse;
 pub const parseText = @import("json_parse.zig").parseText;
 pub const ParseResult = @import("json_parse.zig").Result;
+pub const wtf8Encode = @import("json_parse.zig").wtf8Encode;
 pub const write = @import("json_write.zig").write;
 pub const writeString = @import("json_write.zig").writeString;
 pub const Options = @import("json_write.zig").Options;

zig/src/server/live.zig

@@ -1,7 +1,7 @@
 //! The live status line (connections, decode and prefill tok/s), redrawn in a terminal, and /health's copy of it.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const log = @import("log.zig");
 const Allocator = std.mem.Allocator;
 

zig/src/server/logprobs.zig

@@ -1,7 +1,7 @@
 //! OpenAI ``logprobs`` (who may ask, ``choices[0].logprobs`` from the engine's rows) and vLLM's ``prompt_logprobs``.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const reply_text = @import("reply_text.zig");
 const Server = @import("server.zig").Server;

zig/src/server/main.zig

@@ -1,6 +1,6 @@
 //! ``tensorfold-native``: ``capabilities --json`` for the Python switch, and ``serve MODEL [flags]`` with no Python.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const cli = @import("cli.zig");
 const log = @import("log.zig");
 const hub = @import("hub.zig");

zig/src/server/memory_routes.zig

@@ -1,7 +1,7 @@
 //! Authenticated, read-only sizing diagnostics; unknown counters and plans are omitted rather than guessed.
 const std = @import("std");
 const api = @import("engine_api");
-const json = @import("json.zig");
+const json = @import("json");
 const process = @import("process_memory.zig");
 const routes = @import("routes.zig");
 const Server = @import("server.zig").Server;

zig/src/server/messages.zig

@@ -1,6 +1,6 @@
 //! Chat messages as the template sees them: leading instructions merged, text parts joined, call arguments as objects.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields = @import("fields.zig");
 const Value = json.Value;

zig/src/server/model_text.zig

@@ -1,6 +1,6 @@
 //! The tokenizer and chat template a server renders prompts and replies with (the tok and jinja branches plug in here).
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const Allocator = std.mem.Allocator;
 
 pub const RenderOptions = struct {

zig/src/server/openai.zig

@@ -1,6 +1,6 @@
 //! POST /v1/chat/completions and /v1/completions, JSON or SSE, as ``server.http`` answers them.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields = @import("fields.zig");
 const chat = @import("chat.zig");

zig/src/server/prompt.zig

@@ -1,6 +1,6 @@
 //! Prompt ids from a request, as ``prepare_prompt`` and ``render_prompt_ids`` make them.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const messages_mod = @import("messages.zig");
 const model_text = @import("model_text.zig");

zig/src/server/pyrepr.zig

@@ -1,6 +1,6 @@
 //! Python's ``repr`` of decoded JSON, for refusal messages that quote a request's value.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const Value = json.Value;
 const Allocator = std.mem.Allocator;
 

zig/src/server/request_log.zig

@@ -1,6 +1,6 @@
 //! TENSORFOLD_REQUEST_LOG: each chat and completion body appended as one JSON line, images redacted, for replays.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const Value = json.Value;
 const Allocator = std.mem.Allocator;
 

zig/src/server/responses.zig

@@ -1,6 +1,6 @@
 //! OpenAI's Responses API: each request runs as its chat completion, translated item by item and event by event.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const http_body = @import("http_body.zig");
 const openai = @import("openai.zig");

zig/src/server/responses_include_test.zig

@@ -1,6 +1,6 @@
 //! Responses' include: OpenAI's values accepted and ignored, anything else refused by name.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const responses_translate = @import("responses_translate.zig");
 

zig/src/server/responses_translate.zig

@@ -1,6 +1,6 @@
 //! A Responses request as the chat completion that runs it (``responses_translate.translate``).
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const errors = @import("errors.zig");
 const fields = @import("fields.zig");
 const pyrepr = @import("pyrepr.zig");

zig/src/server/root.zig

@@ -1,5 +1,5 @@
 //! The native HTTP server's surface: what main and the test servers build a server from.
-pub const json = @import("json.zig");
+pub const json = @import("json");
 pub const errors = @import("errors.zig");
 pub const auth = @import("auth.zig");
 pub const cli = @import("cli.zig");
@@ -17,7 +17,7 @@ pub const Config = server.Config;
 
 test {
     @import("std").testing.refAllDecls(@This());
-    _ = @import("json.zig");
+    _ = @import("json");
     _ = @import("std_options.zig");
     _ = @import("fields.zig");
     _ = @import("reply_text.zig");

zig/src/server/routes.zig

@@ -1,6 +1,6 @@
 //! Method and path dispatch, as ``do_GET``, ``do_POST`` and ``do_DELETE`` route them.
 const std = @import("std");
-const json = @import("json.zig");
+const json = @import("json");
 const http_conn = @import("http_conn.zig");
 const http_body = @import("http_body.zig");
 const auth = @import("auth.zig");