Removed unused config values - llm info should only be set in plugin config.

plugin/src/client.ts

@@ -135,11 +135,17 @@ export const runPipeline = async ({
         }),
     })
 
-    const data = await readApiResponse<PipelineRunResponse | { detail?: string }>(response, url)
+    const data = await readApiResponse<PipelineRunResponse | { detail?: unknown }>(response, url)
 
     if (!response.ok) {
+        if (response.status == 422) {
+            throw new Error(
+                'Plugin is not configured. Set the model, API key, and API URL in the plugin settings.',
+            )
+        }
+        const detail = 'detail' in data ? data.detail : undefined
         throw new Error(
-            'detail' in data && data.detail ? data.detail : 'Pipeline execution failed.',
+            typeof detail === 'string' && detail ? detail : 'Pipeline execution failed.',
         )
     }
 

service/config.template.yaml

@@ -1,13 +1,3 @@
-llm_response_generation:
-  type: "openai-api"
-  model: "gpt-oss-120b"
-  api_url: https://llm.ai.e-infra.cz/v1/
-  api_key: "xxx"
-  workers: 4
-
-dsw:
-  api_url: ""
-
 auth:
   allowed_project_urls:
     - "https://your-dsw-instance.example.com"

service/src/ai_document_plugin_service/ai/assignment/assignment_component.py

@@ -24,6 +24,7 @@
 )
 from ai_document_plugin_service.ai.assignment.types import SectionAssignment
 from ai_document_plugin_service.ai.common.config import Config
+from ai_document_plugin_service.ai.common.llm_client import LLMClient
 from ai_document_plugin_service.ai.common.progress import progress_percent
 from ai_document_plugin_service.ai.common.types import AssignmentStats
 from ai_document_plugin_service.ai.knowledgemodel.types import QuestionData
@@ -38,6 +39,17 @@ class AssignmentComponentResult(TypedDict):
 
 @component
 class AssignmentComponent:
+    """
+    This component is responsible for assigning questions from the questionnaire to the sections from the dmp template.
+    For example, it assigns question 'When will the project start?' to sections Introduction and Project Timeline
+    """
+
+    def __init__(self, llm_client: LLMClient, config: Config) -> None:
+        self.llm_client = llm_client
+        self.config = config
+        self.section_id_generator = OpenAISectionIdGenerator(llm_client, config)
+        self.section_matcher = OpenAILayerMatcher(self.llm_client, self.config)
+
     @staticmethod
     def _add_chunk_mapping_to_result(
         *,
@@ -53,16 +65,13 @@ def _add_chunk_mapping_to_result(
                 continue
             result_mapping[question_path] = [section_formatter.record_id_for_sid(sid) for sid in section_ids]
 
-    @staticmethod
     def _match_single_chunk(
-        *,
-        config: Config,
+        self,
         sections_xml: str,
         question_chunk: str,
         stats: AssignmentStats,
     ) -> dict[str, list[str]]:
-        matcher = OpenAILayerMatcher(config)
-        return matcher.match_questions_to_sections(
+        return self.section_matcher.match_questions_to_sections(
             sections_xml,
             question_chunk,
             stats,
@@ -73,7 +82,6 @@ def run(
         self,
         data: list[QuestionData],
         template_data: dict[str, Any],
-        config: Config,
         km: dict[str, Any],
         on_progress: Callable[[str], None] | None = None,
     ) -> AssignmentComponentResult:
@@ -85,7 +93,7 @@ def run(
         stats = AssignmentStats()
 
         section_formatter = SectionFormatter(sections)
-        section_formatter.create_mappings(OpenAISectionIdGenerator(config), stats)
+        section_formatter.create_mappings(self.section_id_generator, stats)
         sections_xml = section_formatter.get_sections_as_xml()
 
         result_mapping = {}
@@ -94,7 +102,6 @@ def run(
 
         def match_chunk(question_chunk: str) -> dict[str, list[str]]:
             result = self._match_single_chunk(
-                config=config,
                 sections_xml=sections_xml,
                 question_chunk=question_chunk,
                 stats=stats,
@@ -109,8 +116,8 @@ def match_chunk(question_chunk: str) -> dict[str, list[str]]:
         for question_to_section_ids in thread_map(
             match_chunk,
             question_chunks,
-            max_workers=config.parallel_workers,
-            desc=f'Assigning questions to sections ({config.parallel_workers} workers)',
+            max_workers=self.llm_client.get_max_workers(),
+            desc=f'Assigning questions to sections ({self.llm_client.get_max_workers()} workers)',
         ):
             self._add_chunk_mapping_to_result(
                 result_mapping=result_mapping,

service/src/ai_document_plugin_service/ai/assignment/llm.py

@@ -45,9 +45,9 @@ def match_questions_to_sections(
 
 
 class OpenAILayerMatcher(LayerMatcher):
-    def __init__(self, config: Config) -> None:
+    def __init__(self, llm_client: LLMClient, config: Config) -> None:
+        self.client = llm_client
         self.config = config
-        self.client = LLMClient(config)
 
     def match_questions_to_sections(
         self,
@@ -74,7 +74,6 @@ def match_questions_to_sections(
 
         def call_and_parse() -> dict[str, list[str]]:
             response = self.client.completion(
-                model=self.config.model,
                 messages=messages,
                 temperature=self.config.assignment.temperature,
                 max_tokens=self.config.assignment.max_tokens,

service/src/ai_document_plugin_service/ai/assignment/sections/llm.py

@@ -1,7 +1,6 @@
 import typing
 from abc import ABC, abstractmethod
 
-from openai import OpenAI
 from tqdm import tqdm
 
 from ai_document_plugin_service.ai.assignment.types import LeafSection
@@ -29,9 +28,9 @@ def generate_leaf_section_ids(
 
 
 class OpenAISectionIdGenerator(SectionIdGenerator):
-    def __init__(self, config: Config) -> None:
+    def __init__(self, llm_client: LLMClient, config: Config) -> None:
         self.config = config
-        self.client = LLMClient(config)
+        self.client = llm_client
 
     @typing.override
     def generate_leaf_section_ids(
@@ -59,7 +58,6 @@ def generate_leaf_section_ids(
             )
             response = call_with_retry(
                 lambda um=user_msg: self.client.completion(
-                    model=self.config.model,
                     messages=[
                         {'role': 'system', 'content': system_msg},
                         {'role': 'user', 'content': um},
@@ -92,7 +90,6 @@ def _normalize_section_id(raw: str) -> str:
 class LoggingNoopSectionIdGenerator(SectionIdGenerator):
     def __init__(self, config: Config) -> None:
         self.config = config
-        self.client = OpenAI(api_key=config.api_key, base_url=config.api_url)
 
     @typing.override
     def generate_leaf_section_ids(  # ty: ignore[invalid-method-override]

service/src/ai_document_plugin_service/ai/common/config.py

@@ -1,6 +1,6 @@
 import os
 import pathlib
-from dataclasses import dataclass, replace
+from dataclasses import dataclass
 from typing import Any
 
 import yaml
@@ -41,26 +41,21 @@ class DatabaseConfig:
 
 @dataclass(frozen=True)
 class Config:
-    api_key: str
-    api_url: str
-    dsw_api_url: str
     allowed_project_urls: tuple[str, ...]
-    model: str
     log_level: str
     database: DatabaseConfig
     files: FilePaths
     assignment: SystemAndUserPrompt
     section_id: SystemAndUserPrompt
     dmp_generation: SystemPrompt
     dmp_polishing: SystemAndUserPrompt
-    parallel_workers: int
 
 
 @dataclass(frozen=True)
-class LLMConfigOverride:
-    model: str | None = None
-    api_key: str | None = None
-    api_url: str | None = None
+class LLMConfig:
+    model: str
+    api_key: str
+    api_url: str
     parallel_workers: int | None = None
 
 
@@ -134,19 +129,6 @@ def _get_allowed_project_urls(config: dict[str, Any]) -> tuple[str, ...]:
     return tuple(normalized)
 
 
-def _get_parallel_workers(config: dict[str, Any]) -> int:
-    workers = config.get('llm_response_generation', {}).get('workers', 1)
-    try:
-        workers_int = int(workers)
-    except (TypeError, ValueError) as exc:
-        msg = "Invalid config value: 'parallelism.workers' must be an integer >= 1"
-        raise ValueError(msg) from exc
-    if workers_int < 1:
-        msg = "Invalid config value: 'parallelism.workers' must be >= 1"
-        raise ValueError(msg)
-    return workers_int
-
-
 def _resolve_existing_path(path: str, *, base_dir: pathlib.Path | None = None) -> str:
     normalized_path = _normalize_path(path)
     path_obj = pathlib.Path(normalized_path)
@@ -199,12 +181,6 @@ def load_config(config_path: str | None = None) -> Config:
         raise TypeError(msg)
 
     return Config(
-        api_key=_expand_env_vars(
-            _get(config, 'llm_response_generation', 'api_key'),
-        ),
-        api_url=_get(config, 'llm_response_generation', 'api_url'),
-        model=_get(config, 'llm_response_generation', 'model'),
-        dsw_api_url=_get(config, 'dsw', 'api_url'),
         allowed_project_urls=_get_allowed_project_urls(config),
         log_level=_get_log_level(config),
         database=DatabaseConfig(
@@ -241,18 +217,4 @@ def load_config(config_path: str | None = None) -> Config:
             system_message=_get(prompts, 'dmp_polishing', 'system_message'),
             user_message=_get(prompts, 'dmp_polishing', 'user_message'),
         ),
-        parallel_workers=_get_parallel_workers(config),
-    )
-
-
-def apply_llm_override(config: Config, override: LLMConfigOverride | None = None) -> Config:
-    if override is None:
-        return config
-
-    return replace(
-        config,
-        model=override.model or config.model,
-        api_key=override.api_key or config.api_key,
-        api_url=override.api_url or config.api_url,
-        parallel_workers=override.parallel_workers or config.parallel_workers,
     )

service/src/ai_document_plugin_service/ai/common/llm_client.py

@@ -7,7 +7,6 @@
 from openai import APIConnectionError, APITimeoutError, OpenAI, RateLimitError
 from openai.types.chat import ChatCompletion
 
-from ai_document_plugin_service.ai.common.config import Config
 from ai_document_plugin_service.ai.common.dynamic_semaphore import DynamicSemaphore
 
 if TYPE_CHECKING:
@@ -76,27 +75,33 @@ def add_usage(stats: 'AssignmentStats | None', response: object) -> None:
 
 
 class LLMClient:
-    def __init__(self, config: Config) -> None:
-        self.client = OpenAI(api_key=config.api_key, base_url=config.api_url, max_retries=0)
-        self.max_workers = config.parallel_workers or 1
+    def __init__(self, model: str, api_key: str, api_url: str, parallel_workers: int | None) -> None:
+        self.client = OpenAI(api_key=api_key, base_url=api_url, max_retries=0)
+        self.model = model
+        self.max_workers = parallel_workers or 1
         logger.debug(
             'Initializing LLM client, setting semaphore limit to %s',
             self.max_workers,
         )
         semaphore.set_limit(self.max_workers)
 
+    def get_max_workers(self) -> int:
+        return self.max_workers
+
+    def get_model_name(self) -> str:
+        return self.model
+
     def completion(
         self,
         *args: Any,  # noqa: ANN401
         **kwargs: Any,  # noqa: ANN401
     ) -> ChatCompletion:
         req_id = uuid.uuid4().hex[:8]
-        model = kwargs.get('model', args[0] if args else '?')
         wait_start = time.perf_counter()
         logger.debug(
             '[llm] req=%s model=%s queueing (semaphore active/limit unknown until acquire)',
             req_id,
-            model,
+            self.model,
         )
         with semaphore:
             wait_s = time.perf_counter() - wait_start
@@ -107,7 +112,7 @@ def completion(
                 semaphore.limit,
             )
             call_start = time.perf_counter()
-            result = self.client.chat.completions.create(*args, **kwargs)
+            result = self.client.chat.completions.create(*args, model=self.model, **kwargs)
             logger.debug(
                 '[llm] req=%s completed in %.3fs (releasing semaphore)',
                 req_id,

service/src/ai_document_plugin_service/ai/generation/dmp_generator_component.py

@@ -11,12 +11,10 @@
 from tqdm import tqdm
 
 from ai_document_plugin_service.ai.assignment.types import SerializedSectionAssignment
-from ai_document_plugin_service.ai.common import Config
 from ai_document_plugin_service.ai.common.progress import progress_percent
 from ai_document_plugin_service.ai.common.types import AssignmentStats
 from ai_document_plugin_service.ai.generation.llm import (
     GenerationLLM,
-    OpenAIGenerationLLM,
 )
 from ai_document_plugin_service.ai.generation.parse_answers import parse_answer
 
@@ -42,16 +40,14 @@ class _ScheduledSection:
 
 @component
 class DmpGeneratorComponent:
-    def __init__(self, llm: GenerationLLM | None = None) -> None:
-        self.llm = llm
+    def __init__(self, dmp_generator_llm: GenerationLLM) -> None:
+        self.dmp_generator_llm = dmp_generator_llm
 
     @component.output_types(markdown=str, debug_markdown=str, stats=AssignmentStats)
     def run(
         self,
         replies: dict,
         km: dict,
-        config: Config,
-        llm: GenerationLLM | None = None,
         new_assignments: list[SerializedSectionAssignment] | None = None,
         db_assignments: list[SerializedSectionAssignment] | None = None,
         on_progress: Callable[[str], None] | None = None,
@@ -66,18 +62,15 @@ def run(
         replies = self._filter_reachable_replies(replies, km)
 
         stats = AssignmentStats()
-
-        active_llm = llm or self.llm or OpenAIGenerationLLM(config)
-
-        worker_count = max(1, config.parallel_workers)
-        with ThreadPoolExecutor(max_workers=worker_count) as executor:
+        max_workers = self.dmp_generator_llm.get_max_workers()
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
             scheduled_sections = [
                 self._schedule_section(
                     node=node,
                     depth=0,
                     replies=replies,
                     km=km,
-                    llm=active_llm,
+                    llm=self.dmp_generator_llm,
                     stats=stats,
                     executor=executor,
                 )
@@ -90,7 +83,7 @@ def run(
             for i, future in tqdm(
                 enumerate(as_completed(leaf_futures), start=1),
                 total=total_sections,
-                desc=f'Generating sections ({worker_count} workers)',
+                desc=f'Generating sections ({max_workers} workers)',
             ):
                 future.result()
                 if on_progress is not None: