import { revisionAtLeast } from "./compat/revision";
import { classifyModel } from "./compat/taxonomy";
import type { ModelIdentity } from "./compat/types";
import { bareModelId } from "./identity/id";
import type { ModelTokenizer } from "./types";

// Resolution is pure and catalog ids recur across providers and aliases.
const MAX_TOKENIZER_CACHE_ENTRIES = 2_048;
const modelTokenizerCache = new Map<string, ModelTokenizer | null>();

function claudeTokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	if (identity.class !== "anthropic") return undefined;
	if (identity.family === "opus") {
		if (revisionAtLeast(identity.revision, "5")) return "claude-v5";
		if (revisionAtLeast(identity.revision, "4.7")) return "claude-v47";
		return "claude-v3";
	}
	if (identity.family === "sonnet" || identity.family === "fable" || identity.family === "mythos") {
		return revisionAtLeast(identity.revision, "5") ? "claude-v5-sonnet" : "claude-v3";
	}
	// Haiku 5.5 moved to the post-4.7 tokenizer (~30% more tokens than Haiku
	// 4.5). No Haiku-measured vocabulary is embedded; Sonnet 5's is the nearest.
	if (identity.family === "haiku" && revisionAtLeast(identity.revision, "5.5")) return "claude-v5-sonnet";
	return "claude-v3";
}

function qwenTokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	return identity.class === "qwen" && revisionAtLeast(identity.revision, "3.5") ? "qwen3" : undefined;
}

function deepSeekTokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	return identity.class === "deepseek" ? "deepseek-v3" : undefined;
}

function kimiTokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	return identity.class === "kimi" ? "kimi-k2" : undefined;
}

function glmTokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	return identity.class === "glm" && revisionAtLeast(identity.revision, "5") ? "glm5" : undefined;
}

// GPT-5 and later bill text in o200k_base, GPT-3.5 and GPT-4 (Turbo) in cl100k_base. GPT-4.1, GPT-4.5, gpt-4o and the
// o-series are o200k too, which is the default for a model without a tokenizer.
function openAITokenizer(identity: ModelIdentity): ModelTokenizer | undefined {
	if (identity.class !== "openai") return undefined;
	if (revisionAtLeast(identity.revision, "5")) return "o200k";
	if (
		identity.family === "gpt" &&
		revisionAtLeast(identity.revision, "3.5") &&
		!revisionAtLeast(identity.revision, "4.1")
	) {
		return "cl100k";
	}
	return undefined;
}

/**
 * Resolve the exact locally embedded tokenizer for a canonical model id.
 *
 * This is catalog policy, not a runtime caller heuristic: [`buildModel`](./build.ts)
 * materializes the result as `Model.tokenizer`; consumers read that property.
 */
export function resolveModelTokenizer(modelId: string, provider = ""): ModelTokenizer | undefined {
	const cacheKey = `${provider}\0${modelId}`;
	const cached = modelTokenizerCache.get(cacheKey);
	if (cached !== undefined) return cached ?? undefined;
	// Provider-scoped identity overrides (e.g. Yolo-Auto's `yolo` alias for
	// Qwen3.8 Flash) only apply when the provider is known; bare ids stay
	// provider-agnostic so unrelated namespaces never inherit them.
	const identity = classifyModel(provider, bareModelId(modelId), { lenient: true });
	const tokenizer =
		claudeTokenizer(identity) ??
		qwenTokenizer(identity) ??
		deepSeekTokenizer(identity) ??
		kimiTokenizer(identity) ??
		glmTokenizer(identity) ??
		openAITokenizer(identity);
	if (modelTokenizerCache.size === MAX_TOKENIZER_CACHE_ENTRIES) modelTokenizerCache.clear();
	modelTokenizerCache.set(cacheKey, tokenizer ?? null);
	return tokenizer;
}