gsd-pi 2.58.0-dev.778d6ac → 2.58.0-dev.d63175c

package/src/resources/extensions/ollama/ollama-client.ts

@@ -1,196 +0,0 @@
-// GSD2 — HTTP client for Ollama REST API
-
-/**
- * Low-level HTTP client for the Ollama REST API.
- * Respects the OLLAMA_HOST environment variable for non-default endpoints.
- *
- * Reference: https://github.com/ollama/ollama/blob/main/docs/api.md
- */
-
-import type {
-	OllamaPsResponse,
-	OllamaPullProgress,
-	OllamaShowResponse,
-	OllamaTagsResponse,
-	OllamaVersionResponse,
-} from "./types.js";
-
-const DEFAULT_HOST = "http://localhost:11434";
-const PROBE_TIMEOUT_MS = 1500;
-const REQUEST_TIMEOUT_MS = 10000;
-
-/**
- * Get the Ollama host URL from OLLAMA_HOST or default.
- */
-export function getOllamaHost(): string {
-	const host = process.env.OLLAMA_HOST;
-	if (!host) return DEFAULT_HOST;
-
-	// OLLAMA_HOST can be just a host:port without scheme
-	if (host.startsWith("http://") || host.startsWith("https://")) return host;
-	return `http://${host}`;
-}
-
-async function fetchWithTimeout(url: string, options: RequestInit = {}, timeoutMs = REQUEST_TIMEOUT_MS): Promise<Response> {
-	const controller = new AbortController();
-	const timeout = setTimeout(() => controller.abort(), timeoutMs);
-	try {
-		return await fetch(url, { ...options, signal: controller.signal });
-	} finally {
-		clearTimeout(timeout);
-	}
-}
-
-/**
- * Check if Ollama is running and reachable.
- */
-export async function isRunning(): Promise<boolean> {
-	try {
-		const response = await fetchWithTimeout(`${getOllamaHost()}/`, {}, PROBE_TIMEOUT_MS);
-		return response.ok;
-	} catch {
-		return false;
-	}
-}
-
-/**
- * Get Ollama version.
- */
-export async function getVersion(): Promise<string | null> {
-	try {
-		const response = await fetchWithTimeout(`${getOllamaHost()}/api/version`);
-		if (!response.ok) return null;
-		const data = (await response.json()) as OllamaVersionResponse;
-		return data.version;
-	} catch {
-		return null;
-	}
-}
-
-/**
- * List all locally available models.
- */
-export async function listModels(): Promise<OllamaTagsResponse> {
-	const response = await fetchWithTimeout(`${getOllamaHost()}/api/tags`);
-	if (!response.ok) {
-		throw new Error(`Ollama /api/tags returned ${response.status}: ${response.statusText}`);
-	}
-	return (await response.json()) as OllamaTagsResponse;
-}
-
-/**
- * Get detailed information about a specific model.
- */
-export async function showModel(name: string): Promise<OllamaShowResponse> {
-	const response = await fetchWithTimeout(`${getOllamaHost()}/api/show`, {
-		method: "POST",
-		headers: { "Content-Type": "application/json" },
-		body: JSON.stringify({ name }),
-	});
-	if (!response.ok) {
-		throw new Error(`Ollama /api/show returned ${response.status}: ${response.statusText}`);
-	}
-	return (await response.json()) as OllamaShowResponse;
-}
-
-/**
- * List currently loaded/running models.
- */
-export async function getRunningModels(): Promise<OllamaPsResponse> {
-	const response = await fetchWithTimeout(`${getOllamaHost()}/api/ps`);
-	if (!response.ok) {
-		throw new Error(`Ollama /api/ps returned ${response.status}: ${response.statusText}`);
-	}
-	return (await response.json()) as OllamaPsResponse;
-}
-
-/**
- * Pull a model with streaming progress.
- * Calls onProgress for each progress update.
- * Returns when the pull is complete.
- */
-export async function pullModel(
-	name: string,
-	onProgress?: (progress: OllamaPullProgress) => void,
-	signal?: AbortSignal,
-): Promise<void> {
-	const response = await fetch(`${getOllamaHost()}/api/pull`, {
-		method: "POST",
-		headers: { "Content-Type": "application/json" },
-		body: JSON.stringify({ name, stream: true }),
-		signal,
-	});
-
-	if (!response.ok) {
-		const text = await response.text();
-		throw new Error(`Ollama /api/pull returned ${response.status}: ${text}`);
-	}
-
-	if (!response.body) {
-		throw new Error("Ollama /api/pull returned no body");
-	}
-
-	const reader = response.body.getReader();
-	const decoder = new TextDecoder();
-	let buffer = "";
-
-	while (true) {
-		const { done, value } = await reader.read();
-		if (done) break;
-
-		buffer += decoder.decode(value, { stream: true });
-		const lines = buffer.split("\n");
-		buffer = lines.pop() ?? "";
-
-		for (const line of lines) {
-			const trimmed = line.trim();
-			if (!trimmed) continue;
-			try {
-				const progress = JSON.parse(trimmed) as OllamaPullProgress;
-				onProgress?.(progress);
-			} catch {
-				// Skip malformed lines
-			}
-		}
-	}
-
-	// Process remaining buffer
-	if (buffer.trim()) {
-		try {
-			const progress = JSON.parse(buffer.trim()) as OllamaPullProgress;
-			onProgress?.(progress);
-		} catch {
-			// Ignore
-		}
-	}
-}
-
-/**
- * Delete a local model.
- */
-export async function deleteModel(name: string): Promise<void> {
-	const response = await fetchWithTimeout(`${getOllamaHost()}/api/delete`, {
-		method: "DELETE",
-		headers: { "Content-Type": "application/json" },
-		body: JSON.stringify({ name }),
-	});
-	if (!response.ok) {
-		const text = await response.text();
-		throw new Error(`Ollama /api/delete returned ${response.status}: ${text}`);
-	}
-}
-
-/**
- * Copy a model to a new name.
- */
-export async function copyModel(source: string, destination: string): Promise<void> {
-	const response = await fetchWithTimeout(`${getOllamaHost()}/api/copy`, {
-		method: "POST",
-		headers: { "Content-Type": "application/json" },
-		body: JSON.stringify({ source, destination }),
-	});
-	if (!response.ok) {
-		const text = await response.text();
-		throw new Error(`Ollama /api/copy returned ${response.status}: ${text}`);
-	}
-}

package/src/resources/extensions/ollama/ollama-commands.ts

@@ -1,248 +0,0 @@
-// GSD2 — Ollama slash commands
-
-/**
- * Registers /ollama slash commands for managing local Ollama models.
- *
- * Commands:
- *   /ollama          — Show status (running?, version, loaded models)
- *   /ollama list     — List all available local models with sizes
- *   /ollama pull     — Pull a model with progress
- *   /ollama remove   — Delete a local model
- *   /ollama ps       — Show running models and resource usage
- */
-
-import type { ExtensionAPI } from "@gsd/pi-coding-agent";
-import { Text } from "@gsd/pi-tui";
-import * as client from "./ollama-client.js";
-import { discoverModels, formatModelForDisplay } from "./ollama-discovery.js";
-import { formatModelSize } from "./model-capabilities.js";
-
-export function registerOllamaCommands(pi: ExtensionAPI): void {
-	pi.registerCommand("ollama", {
-		description: "Manage local Ollama models — list | pull | remove | ps",
-		async handler(args, ctx) {
-			const parts = (args ?? "").trim().split(/\s+/);
-			const subcommand = parts[0] || "status";
-			const modelArg = parts.slice(1).join(" ");
-
-			switch (subcommand) {
-				case "status":
-					return await handleStatus(ctx);
-				case "list":
-				case "ls":
-					return await handleList(ctx);
-				case "pull":
-					return await handlePull(modelArg, ctx);
-				case "remove":
-				case "rm":
-				case "delete":
-					return await handleRemove(modelArg, ctx);
-				case "ps":
-					return await handlePs(ctx);
-				default:
-					ctx.ui.notify(
-						`Unknown subcommand: ${subcommand}. Use: status, list, pull, remove, ps`,
-						"warning",
-					);
-			}
-		},
-	});
-}
-
-async function handleStatus(ctx: any): Promise<void> {
-	const running = await client.isRunning();
-	if (!running) {
-		ctx.ui.notify(
-			"Ollama is not running. Install from https://ollama.com and run 'ollama serve'",
-			"warning",
-		);
-		return;
-	}
-
-	const version = await client.getVersion();
-	const lines: string[] = [];
-	lines.push(`Ollama${version ? ` v${version}` : ""} — running (${client.getOllamaHost()})`);
-
-	// Show loaded models
-	try {
-		const ps = await client.getRunningModels();
-		if (ps.models && ps.models.length > 0) {
-			lines.push("");
-			lines.push("Loaded:");
-			for (const m of ps.models) {
-				const vram = m.size_vram > 0 ? formatModelSize(m.size_vram) + " VRAM" : "CPU";
-				const expiresAt = new Date(m.expires_at);
-				const idleMs = expiresAt.getTime() - Date.now();
-				const idleMin = Math.max(0, Math.floor(idleMs / 60000));
-				lines.push(`  ${m.name}  ${vram}  expires in ${idleMin}m`);
-			}
-		}
-	} catch {
-		// ps endpoint may not be available on older versions
-	}
-
-	// Show available models
-	try {
-		const models = await discoverModels();
-		if (models.length > 0) {
-			lines.push("");
-			lines.push("Available:");
-			for (const m of models) {
-				lines.push(`  ${formatModelForDisplay(m)}`);
-			}
-		} else {
-			lines.push("");
-			lines.push("No models pulled. Use /ollama pull <model> to get started.");
-		}
-	} catch (err) {
-		lines.push("");
-		lines.push(`Error listing models: ${err instanceof Error ? err.message : String(err)}`);
-	}
-
-	await ctx.ui.custom(
-		(tui: any, theme: any, _kb: any, done: (r: undefined) => void) => {
-			const text = new Text(lines.map((l) => theme.fg("fg", l)).join("\n"), 0, 0);
-			setTimeout(() => done(undefined), 0);
-			return text;
-		},
-	);
-}
-
-async function handleList(ctx: any): Promise<void> {
-	const running = await client.isRunning();
-	if (!running) {
-		ctx.ui.notify("Ollama is not running", "warning");
-		return;
-	}
-
-	const models = await discoverModels();
-	if (models.length === 0) {
-		ctx.ui.notify("No models available. Use /ollama pull <model> to download one.", "info");
-		return;
-	}
-
-	const lines = ["Local Ollama models:", ""];
-	for (const m of models) {
-		lines.push(`  ${formatModelForDisplay(m)}`);
-	}
-
-	await ctx.ui.custom(
-		(tui: any, theme: any, _kb: any, done: (r: undefined) => void) => {
-			const text = new Text(lines.map((l) => theme.fg("fg", l)).join("\n"), 0, 0);
-			setTimeout(() => done(undefined), 0);
-			return text;
-		},
-	);
-}
-
-async function handlePull(modelName: string, ctx: any): Promise<void> {
-	if (!modelName) {
-		ctx.ui.notify("Usage: /ollama pull <model> (e.g. /ollama pull llama3.1:8b)", "warning");
-		return;
-	}
-
-	const running = await client.isRunning();
-	if (!running) {
-		ctx.ui.notify("Ollama is not running", "warning");
-		return;
-	}
-
-	ctx.ui.setWidget("ollama-pull", [`Pulling ${modelName}...`]);
-
-	try {
-		let lastPercent = -1;
-		await client.pullModel(modelName, (progress) => {
-			if (progress.total && progress.completed) {
-				const percent = Math.floor((progress.completed / progress.total) * 100);
-				if (percent !== lastPercent) {
-					lastPercent = percent;
-					const completed = formatModelSize(progress.completed);
-					const total = formatModelSize(progress.total);
-					ctx.ui.setWidget("ollama-pull", [
-						`Pulling ${modelName}... ${percent}% (${completed} / ${total})`,
-					]);
-				}
-			} else if (progress.status) {
-				ctx.ui.setWidget("ollama-pull", [`${modelName}: ${progress.status}`]);
-			}
-		});
-
-		ctx.ui.setWidget("ollama-pull", undefined);
-		ctx.ui.notify(`${modelName} pulled successfully`, "success");
-	} catch (err) {
-		ctx.ui.setWidget("ollama-pull", undefined);
-		ctx.ui.notify(
-			`Failed to pull ${modelName}: ${err instanceof Error ? err.message : String(err)}`,
-			"error",
-		);
-	}
-}
-
-async function handleRemove(modelName: string, ctx: any): Promise<void> {
-	if (!modelName) {
-		ctx.ui.notify("Usage: /ollama remove <model>", "warning");
-		return;
-	}
-
-	const running = await client.isRunning();
-	if (!running) {
-		ctx.ui.notify("Ollama is not running", "warning");
-		return;
-	}
-
-	const confirmed = await ctx.ui.confirm(
-		"Delete model",
-		`Are you sure you want to delete ${modelName}?`,
-	);
-
-	if (!confirmed) return;
-
-	try {
-		await client.deleteModel(modelName);
-		ctx.ui.notify(`${modelName} deleted`, "success");
-	} catch (err) {
-		ctx.ui.notify(
-			`Failed to delete ${modelName}: ${err instanceof Error ? err.message : String(err)}`,
-			"error",
-		);
-	}
-}
-
-async function handlePs(ctx: any): Promise<void> {
-	const running = await client.isRunning();
-	if (!running) {
-		ctx.ui.notify("Ollama is not running", "warning");
-		return;
-	}
-
-	try {
-		const ps = await client.getRunningModels();
-		if (!ps.models || ps.models.length === 0) {
-			ctx.ui.notify("No models currently loaded in memory", "info");
-			return;
-		}
-
-		const lines = ["Running models:", ""];
-		for (const m of ps.models) {
-			const vram = m.size_vram > 0 ? formatModelSize(m.size_vram) + " VRAM" : "CPU only";
-			const totalSize = formatModelSize(m.size);
-			const expiresAt = new Date(m.expires_at);
-			const idleMs = expiresAt.getTime() - Date.now();
-			const idleMin = Math.max(0, Math.floor(idleMs / 60000));
-			lines.push(`  ${m.name}  ${totalSize}  ${vram}  expires in ${idleMin}m`);
-		}
-
-		await ctx.ui.custom(
-			(tui: any, theme: any, _kb: any, done: (r: undefined) => void) => {
-				const text = new Text(lines.map((l) => theme.fg("fg", l)).join("\n"), 0, 0);
-				setTimeout(() => done(undefined), 0);
-				return text;
-			},
-		);
-	} catch (err) {
-		ctx.ui.notify(
-			`Failed to get running models: ${err instanceof Error ? err.message : String(err)}`,
-			"error",
-		);
-	}
-}

package/src/resources/extensions/ollama/ollama-discovery.ts

@@ -1,106 +0,0 @@
-// GSD2 — Ollama model discovery and capability detection
-
-/**
- * Discovers locally available Ollama models and enriches them with
- * capability metadata (context window, vision, reasoning) from the
- * known model table and /api/show responses.
- *
- * Returns models in the format expected by pi.registerProvider().
- */
-
-import { listModels, getOllamaHost } from "./ollama-client.js";
-import {
-	estimateContextFromParams,
-	formatModelSize,
-	getModelCapabilities,
-	humanizeModelName,
-} from "./model-capabilities.js";
-import type { OllamaModelInfo } from "./types.js";
-
-export interface DiscoveredOllamaModel {
-	id: string;
-	name: string;
-	reasoning: boolean;
-	input: ("text" | "image")[];
-	cost: { input: number; output: number; cacheRead: number; cacheWrite: number };
-	contextWindow: number;
-	maxTokens: number;
-	/** Raw size in bytes for display purposes */
-	sizeBytes: number;
-	/** Parameter size string from Ollama (e.g. "7B") */
-	parameterSize: string;
-}
-
-const ZERO_COST = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
-
-function enrichModel(info: OllamaModelInfo): DiscoveredOllamaModel {
-	const caps = getModelCapabilities(info.name);
-	const parameterSize = info.details?.parameter_size ?? "";
-
-	// Determine context window: known table > estimate from param size > default
-	const contextWindow =
-		caps.contextWindow ??
-		(parameterSize ? estimateContextFromParams(parameterSize) : 8192);
-
-	// Determine max tokens: known table > fraction of context > default
-	const maxTokens =
-		caps.maxTokens ?? Math.min(Math.floor(contextWindow / 4), 16384);
-
-	// Detect vision from families or known table
-	const hasVision =
-		caps.input?.includes("image") ??
-		(info.details?.families?.some((f) => f === "clip" || f === "mllama") ?? false);
-
-	// Detect reasoning from known table
-	const reasoning = caps.reasoning ?? false;
-
-	return {
-		id: info.name,
-		name: humanizeModelName(info.name),
-		reasoning,
-		input: hasVision ? ["text", "image"] : ["text"],
-		cost: ZERO_COST,
-		contextWindow,
-		maxTokens,
-		sizeBytes: info.size,
-		parameterSize,
-	};
-}
-
-/**
- * Discover all locally available Ollama models with enriched capabilities.
- */
-export async function discoverModels(): Promise<DiscoveredOllamaModel[]> {
-	const tags = await listModels();
-	if (!tags.models || tags.models.length === 0) return [];
-
-	return tags.models.map(enrichModel);
-}
-
-/**
- * Format a discovered model for display in model list.
- */
-export function formatModelForDisplay(model: DiscoveredOllamaModel): string {
-	const parts = [model.id];
-
-	if (model.sizeBytes > 0) {
-		parts.push(`(${formatModelSize(model.sizeBytes)})`);
-	}
-
-	const flags: string[] = [];
-	if (model.reasoning) flags.push("reasoning");
-	if (model.input.includes("image")) flags.push("vision");
-
-	if (flags.length > 0) {
-		parts.push(`[${flags.join(", ")}]`);
-	}
-
-	return parts.join(" ");
-}
-
-/**
- * Build the OpenAI-compat base URL for Ollama.
- */
-export function getOllamaOpenAIBaseUrl(): string {
-	return `${getOllamaHost()}/v1`;
-}