provider-configuration
Configure DeepSearcher providers, provider dictionaries, credentials, optional dependencies, and setup troubleshooting.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- provider-configuration
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Configure DeepSearcher providers, provider dictionaries, credentials, optional dependencies, and setup troubleshooting.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/configuration-workflows.md3.8 KB
- 参考文档references/provider-matrix.md5.6 KB
- 参考文档references/troubleshooting.md2.3 KB
- 脚本scripts/validate_provider_config.py13.3 KB
03 SKILL.md 正文
Provider Configuration
Use this sub-skill when a task is about selecting or diagnosing DeepSearcher providers for LLMs, embeddings, file loaders, web crawlers, or vector databases. DeepSearcher provider strings are exact Python class names, and init_config(config) constructs every configured provider before ingestion or query work begins.
Route here for
- Building a
Configuration(config_path=...)object from defaults or YAML. - Updating providers with
set_provider_config(feature, provider, provider_configs)and checking them withget_provider_config(feature). - Choosing exact feature keys:
llm,embedding,file_loader,web_crawler,vector_db. - Mapping provider names to credential variables, optional packages, and common config dictionaries.
- Diagnosing provider import failures, missing SDKs, missing credentials, FireCrawl
ScrapeOptionserrors, or local Milvus Lite setup failures.
Route elsewhere
- Loading files, crawling websites, chunking, collections, and indexing:
data-ingestion. query,retrieve,DeepSearch,ChainOfRAG,NaiveRAG, and result interpretation:rag-query.- Console command behavior, FastAPI service endpoints, or help output:
cli-and-service. - 2WikiMultiHopQA or benchmark/evaluation workflows:
evaluation.
Reference map
- Provider matrix: exact provider names, feature keys, credentials, optional dependencies, and default provider choices.
- Configuration workflows: Python/YAML setup patterns,
ModuleFactorybehavior,init_configside effects, and safe provider switches. - Troubleshooting: provider-name mismatches, optional dependency errors, FireCrawl and Milvus Lite version constraints, and credential handling.
Safe bundled helper
Use scripts/validate_provider_config.py to validate a DeepSearcher YAML config or the built-in default provider set without calling provider APIs:
python scripts/validate_provider_config.py --config deepsearcher-config.yaml
python scripts/validate_provider_config.py --list-providers
python scripts/validate_provider_config.py --print-example ollama-fastembed-milvus
--check-imports is optional. It imports configured provider classes to expose missing packages or incompatible SDK versions, but it still does not instantiate providers, read credentials, contact networks, or open vector databases.
Operating rules
- Treat provider strings as exact class names.
JiekouAIis valid;Jiekou.AIis not aModuleFactoryprovider string. - Prefer environment variables for credentials. If a config dictionary contains
api_key,token,password, or cloud access keys, redact values in any output. - Keep optional, credentialed, network, or service-backed workflows optional until the user explicitly chooses them and provides credentials/service readiness.
- When switching providers, update every affected feature before
init_config(config), becauseinit_configconstructs all configured providers, including unused defaults such asFireCrawlCrawlerandMilvus. - For local Milvus Lite, use a unique working directory or explicit
uriper concurrent process to avoid locking the default./milvus.db.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).