refactor(ts): remove non-client effect run boundaries

ts/packages/flow/src/model/text-completion/azure-openai.ts

@@ -176,7 +176,22 @@ const makeAzureOpenAIProviderFromClient = (
 export function makeAzureOpenAIProvider(
   config: AzureOpenAIProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(makeAzureOpenAIProviderEffect(config));
+  const resolved = {
+    defaultModel: config.model ?? "gpt-4o",
+    defaultTemperature: config.temperature ?? 0.0,
+    maxOutput: config.maxOutput ?? 4096,
+    apiKey: config.apiKey ?? "",
+    endpoint: config.endpoint ?? "",
+    apiVersion: config.apiVersion ?? "2024-12-01-preview",
+  } satisfies ResolvedAzureOpenAIConfig;
+  return makeAzureOpenAIProviderFromClient(
+    resolved,
+    new AzureOpenAI({
+      apiKey: resolved.apiKey,
+      apiVersion: resolved.apiVersion,
+      endpoint: resolved.endpoint,
+    }),
+  );
 }
 
 export const makeAzureOpenAIProviderEffect = Effect.fn("makeAzureOpenAIProvider")(function*(

ts/packages/flow/src/model/text-completion/claude.ts

@@ -16,7 +16,7 @@ import {
   makeFlowProcessorProgram,
   makeLlmSpecs,
 } from "@trustgraph/base";
-import { Effect, Layer, Redacted } from "effect";
+import { Context, Effect, Layer, Redacted } from "effect";
 import { FetchHttpClient } from "effect/unstable/http";
 import type {
   TextCompletionConfigError,
@@ -69,7 +69,32 @@ const makeClaudeLayer = (apiKey: string) =>
 export function makeClaudeProvider(
   config: ClaudeProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(Effect.scoped(makeClaudeProviderEffect(config)));
+  const resolved = {
+    defaultModel: config.model ?? "claude-sonnet-4-20250514",
+    defaultTemperature: config.temperature ?? 0.0,
+    maxOutput: config.maxOutput ?? 8192,
+    apiKey: config.apiKey ?? "",
+  } satisfies ResolvedClaudeConfig;
+  return makeLanguageModelProvider({
+    provider: "Claude",
+    defaultModel: resolved.defaultModel,
+    defaultTemperature: resolved.defaultTemperature,
+    context: Context.empty(),
+    makeLanguageModel: ({ model, temperature }) =>
+      Effect.scoped(
+        Layer.build(makeClaudeLayer(resolved.apiKey)).pipe(
+          Effect.flatMap((context) =>
+            AnthropicLanguageModel.make({
+              model,
+              config: {
+                max_tokens: resolved.maxOutput,
+                temperature,
+              },
+            }).pipe(Effect.provideContext(context))
+          ),
+        ),
+      ),
+  });
 }
 
 export const makeClaudeProviderEffect = Effect.fn("makeClaudeProvider")(function* (

ts/packages/flow/src/model/text-completion/mistral.ts

@@ -156,7 +156,16 @@ const makeMistralProviderFromClient = (
 export function makeMistralProvider(
   config: MistralProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(makeMistralProviderEffect(config));
+  const resolved = {
+    defaultModel: config.model ?? "ministral-8b-latest",
+    defaultTemperature: config.temperature ?? 0.0,
+    maxOutput: config.maxOutput ?? 4096,
+    apiKey: config.apiKey ?? "",
+  } satisfies ResolvedMistralConfig;
+  return makeMistralProviderFromClient(
+    resolved,
+    new Mistral({ apiKey: resolved.apiKey }),
+  );
 }
 
 export const makeMistralProviderEffect = Effect.fn("makeMistralProvider")(function*(

ts/packages/flow/src/model/text-completion/ollama.ts

@@ -132,7 +132,14 @@ const makeOllamaProviderFromClient = (
 export function makeOllamaProvider(
   config: OllamaProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(makeOllamaProviderEffect(config));
+  const resolved = {
+    defaultModel: config.model ?? "qwen2.5:0.5b",
+    host: config.ollamaUrl ?? "http://localhost:11434",
+  } satisfies ResolvedOllamaConfig;
+  return makeOllamaProviderFromClient(
+    resolved,
+    new Ollama({ host: resolved.host }),
+  );
 }
 
 export const makeOllamaProviderEffect = Effect.fn("makeOllamaProvider")(function*(

ts/packages/flow/src/model/text-completion/openai-compatible.ts

@@ -165,7 +165,17 @@ const makeOpenAICompatibleProviderFromClient = (
 export function makeOpenAICompatibleProvider(
   config: OpenAICompatibleProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(makeOpenAICompatibleProviderEffect(config));
+  const resolved = {
+    defaultModel: config.model ?? "default",
+    defaultTemperature: config.temperature ?? 0.0,
+    maxOutput: config.maxOutput ?? 4096,
+    apiKey: config.apiKey ?? "sk-no-key-required",
+    baseURL: config.baseUrl ?? "http://localhost:1234/v1",
+  } satisfies ResolvedOpenAICompatibleConfig;
+  return makeOpenAICompatibleProviderFromClient(
+    resolved,
+    new OpenAI({ baseURL: resolved.baseURL, apiKey: resolved.apiKey }),
+  );
 }
 
 export const makeOpenAICompatibleProviderEffect = Effect.fn("makeOpenAICompatibleProvider")(function*(

ts/packages/flow/src/model/text-completion/openai.ts

@@ -155,7 +155,20 @@ const makeOpenAIProviderFromClient = (
 export function makeOpenAIProvider(
   config: OpenAIProcessorConfig,
 ): LlmProvider<TextCompletionRuntimeError> {
-  return Effect.runSync(makeOpenAIProviderEffect(config));
+  const resolved = {
+    defaultModel: config.model ?? "gpt-4o",
+    defaultTemperature: config.temperature ?? 0.0,
+    maxOutput: config.maxOutput ?? 4096,
+    apiKey: config.apiKey ?? "",
+    baseURL: config.baseUrl,
+  } satisfies ResolvedOpenAIConfig;
+  return makeOpenAIProviderFromClient(
+    resolved,
+    new OpenAI({
+      apiKey: resolved.apiKey,
+      baseURL: resolved.baseURL,
+    }),
+  );
 }
 
 export const makeOpenAIProviderEffect = Effect.fn("makeOpenAIProvider")(function*(