Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
This report summarizes scanner evidence for JustVugg/colibri, a C project with 41,127 GitHub stars. SynthScan examined 394,166 lines of code across 1007 source files, recording 3159 pattern matches distributed across 16 syntactic categories. The legacy adjusted pattern signal is 9.3 (Moderate pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 2753 finding(s).
Are imports declared and package/API names verified? 90 finding(s).
Are wrappers, guards, and repeated structures necessary? 2515 finding(s).
Which changes need focused human review? 3159 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 3159 distinct pattern matches across 16 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | c/v4_dsml.py | 104 | def tool_calls_from_openai_format(tool_calls): | STRING |
| LOW | c/v41_dsml.py | 153 | def tool_calls_from_openai_format(tool_calls): | CODE |
| LOW | c/openai_server.py | 746 | def parse_glm_tool_calls_strict(reply, tools): | CODE |
| LOW | c/openai_server.py | 1031 | def parse_arch_tool_calls_spans(reply, tools, tool_reply=None): | CODE |
| LOW | c/openai_server.py | 3392 | def resolve_generation_prompt(messages, body): | CODE |
| LOW | c/openai_server.py | 3627 | def split_thinking_reply_spans(text, enable_thinking=True, add_generation_prompt=True): | CODE |
| LOW | c/openai_server.py | 4255 | def _require_logprobs_for_echo(wanted): | CODE |
| LOW | c/openai_server.py | 4435 | def _refuse_duplicate_candidate(label, seen, requested, param): | CODE |
| LOW | c/openai_server.py | 4457 | def _completions_logprobs_object(prompt_records, generated_records, display_k, texts=None): | CODE |
| LOW | c/openai_server.py | 4618 | def _completion_choice_fields(stats, *, raw_text, text, chat, engine_k, echo, display_k, | CODE |
| LOW | c/openai_server.py | 5142 | def systemone_decision_record(body, form=None): | CODE |
| LOW | c/setup_hw.py | 60 | def memory_from_windows_status(total_phys, avail_phys, avail_pagefile=None): | CODE |
| LOW | c/setup_hw.py | 162 | def cpu_features_from_windows(is_present): | CODE |
| LOW | c/setup_hw.py | 793 | def parse_win_video_controllers(text): | CODE |
| LOW | c/resource_plan.py | 268 | def _vk_released_tensor_bytes(tensor, family, env): | CODE |
| LOW | c/resource_plan.py | 1847 | def _decode_cgroup_membership_path(field, description): | CODE |
| LOW | c/resource_plan.py | 2038 | def _resolve_cgroup_hierarchy(mounts, member, *, v1=False): | CODE |
| LOW | c/resource_plan.py | 2057 | def _cgroup_hierarchy_remaining(views, member, limit_name, current_name, v1=False): | CODE |
| LOW | c/resource_plan.py | 2567 | def _discover_amd_gpus_windows(): | CODE |
| LOW | c/resource_plan.py | 2922 | def _family_expert_cache_knob(family_id, cache_bytes): | CODE |
| LOW | c/family_registry.py | 987 | def _individual_expert_inventory(pattern): | CODE |
| LOW | c/family_registry.py | 1037 | def _inkling_expert_inventory(name, size, config, _dtype=None): | CODE |
| LOW | c/family_registry.py | 1921 | def _resolve_decision_checkpoint(model): | CODE |
| LOW | c/family_registry.py | 2136 | def fixed_resident_contribution(resolved, name, size, dtype=None): | CODE |
| LOW | c/tools/qpack_http_install.py | 244 | def _validate_mirror_base_url(base_url: str) -> str: | CODE |
| LOW | c/tools/qpack_http_install.py | 435 | def _declared_auxiliary_names(plan): | CODE |
| LOW | c/tools/qpack_http_install.py | 698 | def _verify_or_store_prefetched_aux(session, state, name, raw, completed): | CODE |
| LOW | c/tools/qpack_http_install.py | 801 | def _verify_completed_file_state(session, state, file, chunk_size, | CODE |
| LOW | c/tools/eval_glm.py | 83 | def _checked_engine_text_size(length, label): | CODE |
| LOW | c/tools/prepare_dsv41.py | 31 | def build_compressed_token_map(tokenizer): | CODE |
| LOW | c/tools/datapoint.py | 278 | def _measure_persistent_request(engine, prompt, max_new): | CODE |
| LOW | c/tools/repack_fp8_passthrough.py | 579 | def _canonicalize_metadata_order(path): | CODE |
| LOW | c/tools/repack_fp8_passthrough.py | 661 | def _stamps_from_other_passes(outdir, prog_name): | CODE |
| LOW | c/tools/repack_fp8_passthrough.py | 965 | def _copy_standalone_load_files(indir, outdir): | CODE |
| LOW | c/tools/check_ablate_evidence.py | 78 | def _checked_engine_text_size(length, label): | CODE |
| LOW | c/tools/make_glm53_tiny.py | 154 | def initialize_contract_weights(torch, model, head) -> None: | CODE |
| LOW | c/tools/make_glm53_tiny.py | 519 | def write_sparse_attention_case(torch, model, output: Path) -> None: | CODE |
| LOW | c/tools/qpack_install_policy.py | 90 | def _object_without_duplicates(pairs): | CODE |
| LOW | c/tools/mw_ur3_raw_adapter.py | 574 | def parse_request_capture_bytes(raw: bytes, label: str, *, group: bool, | CODE |
| LOW | c/tools/mw_ur3_raw_adapter.py | 666 | def validate_sequential_requests(sids: dict, corpus: dict, | CODE |
| LOW | c/tools/mw_ur3_raw_adapter.py | 981 | def validate_payload_attribution(manifest: dict, tokenizer_raw: bytes, | CODE |
| LOW | c/tools/make_deepseek_v4_tiny.py | 298 | def initialize_router_coverage(torch, model) -> None: | CODE |
| LOW | c/tools/rans_format.py | 291 | def _py_decode_stream_checked(buf, n, freq, start, slot_to_symbol, scale_bits): | CODE |
| LOW | c/tests/test_segment_conformance_manifest.py | 20 | def test_is_all_registered_families_gate(self): | CODE |
| LOW | c/tests/test_segment_conformance_manifest.py | 32 | def test_every_entry_is_an_honest_contract_fixture(self): | CODE |
| LOW | c/tests/test_segment_conformance_manifest.py | 46 | def test_c_fixture_matrix_matches_manifest(self): | CODE |
| LOW | c/tests/test_qwen38_brio.py | 108 | def test_stale_pin_after_unrelated_generation(self): | CODE |
| LOW | c/tests/test_qwen38_brio.py | 124 | def test_shorter_branch_drops_old_tail(self): | CODE |
| LOW | c/tests/test_qwen38_brio.py | 134 | def test_nested_pins_and_sibling_branches(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 152 | def test_integrated_vulkan_gpu(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 157 | def test_nvidia_with_toolkit_takes_cuda(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 163 | def test_nvidia_without_toolkit_falls_to_vulkan_and_says_how(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 170 | def test_engine_without_cuda_path_uses_vulkan(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 176 | def test_integrated_gpu_only_for_engines_measured_faster_there(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 188 | def test_integrated_gpu_when_vulkan_is_asked_for(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 193 | def test_discrete_gpu_runs_vulkan_for_every_engine(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 198 | def test_vulkan_gpu_without_headers_prints_the_package_command(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 373 | def test_an_nvcc_without_the_option_lists_nothing(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 378 | def test_the_release_table_matches_what_nvcc_lists(self): | CODE |
| LOW⚡ | c/tests/test_setup_flow.py | 387 | def test_the_version_answers_when_the_list_does_not(self): | CODE |
| 2119 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | c/download_fp8.py | 149 | except Exception: pass | CODE |
| HIGH | c/term_image.py | 182 | except (ImportError, OSError, ValueError, AttributeError): | CODE |
| HIGH | c/term_image.py | 72 | except (OSError, ValueError): | CODE |
| HIGH | c/openai_server.py | 6308 | except OSError: | CODE |
| HIGH | c/openai_server.py | 6377 | except OSError: | CODE |
| HIGH | c/openai_server.py | 6480 | except Exception: | CODE |
| HIGH | c/openai_server.py | 6704 | except ClientCancelled: | CODE |
| HIGH | c/openai_server.py | 6706 | except ConnectionError: | CODE |
| HIGH | c/openai_server.py | 8267 | except KeyboardInterrupt: | CODE |
| HIGH | c/openai_server.py | 2369 | except (TypeError, ValueError): | CODE |
| HIGH | c/openai_server.py | 5970 | except (OSError, ValueError, AttributeError): | CODE |
| HIGH | c/openai_server.py | 5974 | except subprocess.TimeoutExpired: | CODE |
| HIGH | c/openai_server.py | 841 | except (json.JSONDecodeError, TypeError, ValueError): | CODE |
| HIGH | c/openai_server.py | 6713 | except OSError: | CODE |
| HIGH | c/openai_server.py | 7707 | except OSError: | CODE |
| HIGH | c/openai_server.py | 5990 | except subprocess.TimeoutExpired: | CODE |
| HIGH | c/image_engine.py | 824 | except OSError: | CODE |
| HIGH | c/image_engine.py | 217 | except OSError: | CODE |
| HIGH | c/image_engine.py | 410 | except (OSError, ValueError): | CODE |
| HIGH | c/image_engine.py | 150 | except (OSError, ValueError): | CODE |
| HIGH | c/image_engine.py | 166 | except subprocess.TimeoutExpired: | CODE |
| HIGH | c/image_engine.py | 433 | except (OSError, ValueError): | CODE |
| HIGH | c/image_engine.py | 172 | except (OSError, ValueError): | CODE |
| HIGH | c/image_engine.py | 379 | except ImageEngineError: | CODE |
| HIGH | c/image_engine.py | 422 | except subprocess.TimeoutExpired: | CODE |
| HIGH | c/setup_hw.py | 102 | except (OSError, AttributeError, ImportError, ValueError): | CODE |
| HIGH | c/setup_hw.py | 124 | except Exception: # a malformed cgroup is reported by `coli plan`, not here | CODE |
| HIGH | c/setup_hw.py | 174 | except (OSError, ImportError): | CODE |
| HIGH | c/setup_hw.py | 181 | except (OSError, AttributeError): | CODE |
| HIGH | c/setup_hw.py | 692 | except ValueError: | CODE |
| HIGH | c/setup_hw.py | 697 | except ValueError: | CODE |
| HIGH | c/setup_flow.py | 145 | except OSError: | CODE |
| HIGH | c/setup_flow.py | 1362 | except OSError: | CODE |
| HIGH | c/autotune.py | 85 | except OSError: | CODE |
| HIGH | c/autotune.py | 172 | except FileNotFoundError: | CODE |
| HIGH | c/doctor.py | 380 | except (OSError, UnicodeDecodeError, json.JSONDecodeError, ValueError): | CODE |
| HIGH | c/resource_plan.py | 1488 | except (OSError, ValueError): | CODE |
| HIGH | c/resource_plan.py | 1635 | except (OSError, TypeError, ValueError): | CODE |
| HIGH | c/resource_plan.py | 2797 | except (OSError, ValueError, subprocess.SubprocessError): | CODE |
| HIGH | c/resource_plan.py | 2244 | except OSError: | CODE |
| HIGH | c/resource_plan.py | 2261 | except (OSError, subprocess.SubprocessError, ValueError): | CODE |
| HIGH | c/resource_plan.py | 2268 | except (OSError, subprocess.SubprocessError, ValueError): | CODE |
| HIGH | c/resource_plan.py | 2344 | except OSError: | CODE |
| HIGH | c/resource_plan.py | 2714 | except (OSError, ValueError, subprocess.SubprocessError): | CODE |
| HIGH | c/resource_plan.py | 1633 | except OSError: | CODE |
| HIGH | c/mcp_server.py | 431 | except (AttributeError, OSError, ValueError): | CODE |
| HIGH | c/tools/convert_qwen38_experts_int4.py | 75 | except AttributeError: | CODE |
| HIGH | c/tools/convert_inkling_dense_int4.py | 50 | except AttributeError: | CODE |
| HIGH | c/tools/iq3_pack.py | 93 | except (OSError, AttributeError): | CODE |
| HIGH | c/tools/qwenimage_stub.py | 233 | except ValueError: | CODE |
| HIGH | c/tools/diag_harness.py | 144 | except: pass | CODE |
| HIGH | c/tools/diag_harness.py | 156 | except: pass | CODE |
| HIGH | c/tools/diag_harness.py | 139 | except (ValueError, IndexError): pass | CODE |
| HIGH | c/tools/diag_harness.py | 168 | except: pass | CODE |
| HIGH | c/tools/diag_harness.py | 242 | except Exception: pass | CODE |
| HIGH | c/tools/convert_olmoe_merged.py | 211 | except Exception: | CODE |
| HIGH | c/tools/make_qwen38_tiny.py | 32 | except (AttributeError, OSError): | CODE |
| HIGH | c/tools/eval_glm.py | 673 | except FileNotFoundError: pass | CODE |
| HIGH | c/tools/repack_rans.py | 336 | except (KeyError, ValueError): | CODE |
| HIGH | c/tools/convert_qwen36.py | 767 | except Exception: | CODE |
| 59 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | start-here.sh | 1 | #!/bin/sh | COMMENT |
| LOW | docs/benchmarks.md | 41 | # 1) measure YOUR disk the way the engine uses it (parallel 19 MB random reads): | COMMENT |
| LOW | .github/workflows/release.yml | 41 | # arm64 macOS joined COLI_V4_SUPPORTED; the archive verify step | COMMENT |
| LOW | .github/workflows/release.yml | 141 | COMMENT | |
| LOW | .github/workflows/release.yml | 181 | # it. Without this the engine is in the archive but `coli chat` still cannot | COMMENT |
| LOW | .github/workflows/release.yml | 241 | # script in the archive the engine ships but cannot be driven by coli chat. | COMMENT |
| LOW | .github/workflows/release.yml | 521 | if [ -z "$NOTES" ]; then | COMMENT |
| LOW | .github/workflows/ci.yml | 21 | run: cd c && make colibri inkling | COMMENT |
| LOW | .github/workflows/ci.yml | 61 | # docker/Dockerfile is not built here (see above) but it is still shipped | COMMENT |
| LOW | .github/workflows/ci.yml | 81 | name: Sanitizers (ASan + UBSan) | COMMENT |
| LOW | .github/workflows/ci.yml | 601 | cd c | COMMENT |
| LOW | .github/workflows/ci.yml | 641 | # (observed: 47 min vs a 5-6 min normal run), so the test harness | COMMENT |
| LOW | .github/workflows/ci.yml | 761 | - name: Install torch (CPU) + transformers | COMMENT |
| LOW | .github/workflows/ci.yml | 781 | # flip a greedy argmax here -- measured. The DENSE weights are the ones | COMMENT |
| LOW | .github/workflows/ci.yml | 901 | # the output. The sanitizers are the half of this gate that watches | COMMENT |
| LOW | .github/workflows/ci.yml | 1521 | # nothing else. A wrong reuse length does not crash -- it answers from | COMMENT |
| LOW | .github/workflows/ci.yml | 1601 | python3 tools/convert_fp8_to_int4.py --indir glm_tiny --outdir glm_tiny_i4 --ebits 4 --io-bits 4 | COMMENT |
| LOW | .github/workflows/ci.yml | 1661 | # link of the test binary below covers the single-source claim, and | COMMENT |
| LOW | .github/workflows/ci.yml | 1681 | echo "HIP syntax check passed" | COMMENT |
| LOW | .github/workflows/ci.yml | 1701 | # windows-2022, NOT windows-latest: the latest image now ships Visual Studio | COMMENT |
| LOW | c/olmoe.c | 21 | * ROUTE_TRACE=<path>: log every routing decision (one line per moe call, | COMMENT |
| LOW | c/olmoe.c | 41 | #endif | COMMENT |
| LOW | c/exact_dot.h | 21 | * Author: Ry Bruscoe. */ | COMMENT |
| LOW | c/inkling.c | 21 | * of the fused [E, 2I, D] / [E, D, I] tensors, LRU-cached, optionally | COMMENT |
| LOW | c/inkling.c | 41 | #include "omp_tune.h" | COMMENT |
| LOW | c/st.h | 1 | /* Indicizzazione e lettura on-demand di tensori da piu' file safetensors. | COMMENT |
| LOW | c/vk_tier.c | 21 | * to the CPU and the tier goes on with the primary one. */ | COMMENT |
| LOW | c/uring.h | 1 | #ifndef COLI_URING_H | COMMENT |
| LOW | c/qwenimage.c | 21 | * qwenimage --model DIR --prompt "..." [--width W] [--height H] [--steps N] | COMMENT |
| LOW | c/qwenimage.c | 41 | #include "json.h" | COMMENT |
| LOW | c/quant.h | 1 | /* quant.h — quantized matmul kernels (header-only, all functions static). | COMMENT |
| LOW | c/qwen36.c | 21 | #define _GNU_SOURCE | COMMENT |
| LOW | c/qwen36.c | 61 | int v = (e && *e) ? atoi(e) : g_q36_default_ctx; | COMMENT |
| LOW | c/qwen36.c | 81 | #include "vk_tier.h" /* optional Vulkan routed-expert tier (COLI_VULKAN=1; stubs without VK=1) */ | COMMENT |
| LOW | c/qwen36.c | 121 | * the same cached choice as q36_walloc, including an int8 copy discarded | COMMENT |
| LOW | c/rans.h | 81 | #error "rans.h assumes a little-endian host (record integers are little-endian)" | COMMENT |
| LOW | c/backend_vulkan.c | 1 | // Vulkan compute backend for colibri's quantized matmul, targeting the | COMMENT |
| LOW | c/backend_vulkan.c | 21 | #endif | COMMENT |
| LOW | c/qi_gemm.h | 21 | #ifndef COLIBRI_QI_GEMM_H | COMMENT |
| LOW | c/vk_chain.h | 41 | * tests/test_vk_chain.c checks every op against a CPU reference (make vk-chain-check). */ | COMMENT |
| LOW | c/vk_chain.h | 721 | #endif | COMMENT |
| LOW | c/gliner_decide.c | 61 | #include <stdint.h> | COMMENT |
| LOW | c/qwen38.c | 21 | #endif | COMMENT |
| LOW | c/qwen38.c | 61 | #include <unistd.h> | COMMENT |
| LOW | c/kimi_k3.c | 81 | * COLI_K3_CKPT_DIR=dir park the photos in files under dir instead of RAM | COMMENT |
| LOW | c/kimi_k3.c | 101 | #include <pthread.h> | COMMENT |
| LOW | c/kimi_k3.c | 121 | #include "pin_pool.h" /* coli_pin_slots_wanted: quanti scatti tenere */ | COMMENT |
| LOW | c/serve_poll.h | 21 | * stdin e' pronto, si risponde "non pronto" e il comportamento torna quello di | COMMENT |
| LOW | c/deepseek_v4.c | 1141 | #endif /* COLI_V4_UNIT_LAYER_RESIDENT */ | COMMENT |
| LOW | c/deepseek_v4.c | 1941 | COMMENT | |
| LOW | c/deepseek_v4.c | 2441 | #define coli_v4_attention_window_token_ref coli_v4_attention_window_token_batch_serial_copy | COMMENT |
| LOW | c/deepseek_v4.c | 3721 | #ifdef COLI_V4_UNIT_INDEXER | COMMENT |
| LOW | c/deepseek_v4.c | 4561 | * COLI_V4_DISABLE_DUAL_EXPERT_LOADER restores the one-worker pipeline. */ | COMMENT |
| LOW | c/deepseek_v4.c | 6801 | #endif /* COLI_V4_UNIT_BLOCK_HYBRID */ | COMMENT |
| LOW | c/deepseek_v4.c | 7201 | void coli_v4_compressor_snapshot_destroy(ColiV4CompressorSnapshot *snapshot) { | COMMENT |
| LOW | c/deepseek_v4.c | 12701 | } | COMMENT |
| LOW | c/deepseek_v4.c | 12721 | #include <float.h> | COMMENT |
| LOW | c/deepseek_v4.c | 17321 | /* ######## deepseek_v4_expert_store.c ######## */ | COMMENT |
| LOW | c/deepseek_v4.c | 17901 | #endif /* COLI_V4_UNIT_EXPERT_STORE */ | COMMENT |
| LOW | c/deepseek_v4.c | 18441 | COMMENT | |
| 184 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | web/tests/ui_migration.py | 35 | CODE | |
| LOW | c/setup_download.py | 176 | CODE | |
| LOW | c/setup_download.py | 293 | CODE | |
| LOW | c/setup_download.py | 334 | CODE | |
| LOW | c/download_fp8.py | 94 | CODE | |
| LOW | c/download_fp8.py | 183 | CODE | |
| LOW | c/v41_dsml.py | 172 | CODE | |
| LOW | c/term_image.py | 40 | CODE | |
| LOW | c/term_image.py | 473 | CODE | |
| LOW | c/openai_server.py | 621 | CODE | |
| LOW | c/openai_server.py | 810 | CODE | |
| LOW | c/openai_server.py | 955 | CODE | |
| LOW | c/openai_server.py | 1326 | CODE | |
| LOW | c/openai_server.py | 1432 | CODE | |
| LOW | c/openai_server.py | 1464 | CODE | |
| LOW | c/openai_server.py | 1602 | CODE | |
| LOW | c/openai_server.py | 1717 | CODE | |
| LOW | c/openai_server.py | 1779 | CODE | |
| LOW | c/openai_server.py | 2011 | CODE | |
| LOW | c/openai_server.py | 2042 | CODE | |
| LOW | c/openai_server.py | 2138 | CODE | |
| LOW | c/openai_server.py | 2227 | CODE | |
| LOW | c/openai_server.py | 2261 | CODE | |
| LOW | c/openai_server.py | 2509 | CODE | |
| LOW | c/openai_server.py | 2591 | CODE | |
| LOW | c/openai_server.py | 2825 | CODE | |
| LOW | c/openai_server.py | 2883 | CODE | |
| LOW | c/openai_server.py | 2924 | CODE | |
| LOW | c/openai_server.py | 3047 | CODE | |
| LOW | c/openai_server.py | 3221 | CODE | |
| LOW | c/openai_server.py | 3278 | CODE | |
| LOW | c/openai_server.py | 3701 | CODE | |
| LOW | c/openai_server.py | 3811 | CODE | |
| LOW | c/openai_server.py | 4096 | CODE | |
| LOW | c/openai_server.py | 4551 | CODE | |
| LOW | c/openai_server.py | 2155 | CODE | |
| LOW | c/openai_server.py | 3586 | CODE | |
| LOW | c/openai_server.py | 5448 | CODE | |
| LOW | c/openai_server.py | 5663 | CODE | |
| LOW | c/openai_server.py | 5721 | CODE | |
| LOW | c/openai_server.py | 5953 | CODE | |
| LOW | c/openai_server.py | 6575 | CODE | |
| LOW | c/openai_server.py | 6663 | CODE | |
| LOW | c/openai_server.py | 6923 | CODE | |
| LOW | c/openai_server.py | 7345 | CODE | |
| LOW | c/openai_server.py | 7812 | CODE | |
| LOW | c/image_engine.py | 610 | CODE | |
| LOW | c/image_engine.py | 767 | CODE | |
| LOW | c/image_engine.py | 117 | CODE | |
| LOW | c/image_engine.py | 180 | CODE | |
| LOW | c/image_engine.py | 232 | CODE | |
| LOW | c/image_engine.py | 263 | CODE | |
| LOW | c/image_engine.py | 282 | CODE | |
| LOW | c/image_engine.py | 405 | CODE | |
| LOW | c/setup_hw.py | 186 | CODE | |
| LOW | c/setup_hw.py | 656 | CODE | |
| LOW | c/setup_flow.py | 500 | CODE | |
| LOW | c/setup_flow.py | 1504 | CODE | |
| LOW | c/autotune.py | 242 | CODE | |
| LOW | c/autotune.py | 413 | CODE | |
| 205 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | c/v4_dsml.py | 96 | except Exception: | STRING |
| LOW | c/v4_dsml.py | 121 | except Exception: | CODE |
| MEDIUM | c/v4_dsml.py | 93 | def to_json(value): | CODE |
| LOW | c/download_fp8.py | 118 | except Exception as e: | CODE |
| LOW | c/download_fp8.py | 149 | except Exception: pass | CODE |
| LOW | c/download_fp8.py | 226 | except Exception as e: | CODE |
| LOW⚡ | c/v41_dsml.py | 115 | except Exception: | STRING |
| LOW | c/v41_dsml.py | 184 | except Exception: | CODE |
| MEDIUM | c/v41_dsml.py | 112 | def to_json(value): | CODE |
| MEDIUM | c/openai_server.py | 1225 | def _np(): | CODE |
| MEDIUM | c/openai_server.py | 5448 | def _dispatch_stdout(self): | CODE |
| MEDIUM | c/openai_server.py | 6305 | def setup(self): | CODE |
| MEDIUM | c/openai_server.py | 7711 | def client_disconnected(self): | CODE |
| LOW | c/openai_server.py | 1356 | except Exception: | CODE |
| LOW | c/openai_server.py | 2749 | except Exception: | CODE |
| LOW | c/openai_server.py | 4884 | except Exception: # not a checkpoint this can read: nothing to decide | CODE |
| LOW | c/openai_server.py | 4896 | except Exception as error: # the engine's own default stands | CODE |
| LOW | c/openai_server.py | 5026 | except Exception: | CODE |
| LOW | c/openai_server.py | 5378 | except Exception as error: | CODE |
| LOW | c/openai_server.py | 5643 | except Exception as error: | CODE |
| LOW | c/openai_server.py | 5657 | except Exception: | CODE |
| LOW | c/openai_server.py | 5699 | except Exception: | CODE |
| LOW | c/openai_server.py | 5824 | except Exception: | CODE |
| LOW | c/openai_server.py | 6128 | except Exception: | CODE |
| LOW | c/openai_server.py | 6480 | except Exception: | CODE |
| LOW | c/openai_server.py | 6708 | except Exception as error: | CODE |
| LOW | c/openai_server.py | 8215 | except Exception: | CODE |
| LOW | c/openai_server.py | 8221 | except Exception: | CODE |
| LOW | c/image_engine.py | 256 | except Exception as error: # the stream is unusable from here on | CODE |
| MEDIUM | c/image_engine.py | 214 | def _kill_quietly(self): | CODE |
| LOW | c/setup_hw.py | 124 | except Exception: # a malformed cgroup is reported by `coli plan`, not here | CODE |
| LOW | c/setup_hw.py | 207 | except Exception: | CODE |
| LOW | c/setup_hw.py | 930 | except Exception as error: # report, never crash the parent's parse | CODE |
| LOW | c/setup_hw.py | 936 | except Exception as error: # same rule as the Vulkan probe | CODE |
| MEDIUM | c/setup_hw.py | 87 | def _windows_memory(): | CODE |
| LOW | c/setup_flow.py | 1033 | except Exception: | CODE |
| LOW | c/setup_flow.py | 1665 | except Exception as error: # the plan is advice here; `coli plan` reports it fully | CODE |
| LOW | c/setup_flow.py | 1695 | except Exception: | CODE |
| LOW | c/setup_flow.py | 1767 | except Exception as error: | CODE |
| MEDIUM | c/setup_flow.py | 119 | def _read_json(path): | CODE |
| MEDIUM | c/setup_flow.py | 1210 | def _get_json(url, timeout=2.0): | CODE |
| LOW | c/resource_plan.py | 98 | except Exception: | CODE |
| LOW | c/resource_plan.py | 132 | except Exception: | CODE |
| LOW | c/mcp_server.py | 409 | except Exception as error: # a tool bug must not take the server down | CODE |
| LOW | c/tools/qpack_http_install.py | 377 | except Exception: | CODE |
| LOW | c/tools/diag_harness.py | 217 | except Exception as e: | CODE |
| LOW | c/tools/diag_harness.py | 242 | except Exception: pass | CODE |
| LOW | c/tools/diag_harness.py | 309 | except Exception as e: | CODE |
| LOW | c/tools/diag_harness.py | 519 | except Exception as e: | CODE |
| LOW | c/tools/mirror_plan.py | 287 | except Exception: | CODE |
| LOW | c/tools/mirror_plan.py | 307 | except Exception: | CODE |
| MEDIUM | c/tools/mirror_plan.py | 39 | def load_json(path): | CODE |
| LOW | c/tools/benchmark_http_serving.py | 167 | except Exception as exc: | CODE |
| LOW | c/tools/convert_olmoe_merged.py | 211 | except Exception: | CODE |
| LOW | c/tools/eval_glm.py | 345 | except Exception: mt = "" | CODE |
| LOW | c/tools/convert_qwen36.py | 296 | except Exception: | CODE |
| LOW | c/tools/convert_qwen36.py | 335 | except Exception as e: | CODE |
| LOW | c/tools/convert_qwen36.py | 345 | except Exception: | CODE |
| LOW | c/tools/convert_qwen36.py | 498 | except Exception: | CODE |
| LOW | c/tools/convert_qwen36.py | 531 | except Exception as e: | CODE |
| 80 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | c/v4_dsml.py | 36 | # ============================================================ | COMMENT |
| MEDIUM | c/v4_dsml.py | 38 | # ============================================================ | COMMENT |
| MEDIUM⚡ | c/v4_dsml.py | 89 | # ============================================================ | STRING |
| MEDIUM⚡ | c/v4_dsml.py | 91 | # ============================================================ | STRING |
| MEDIUM | c/v4_dsml.py | 168 | # ============================================================ | COMMENT |
| MEDIUM | c/v4_dsml.py | 170 | # ============================================================ | COMMENT |
| MEDIUM | c/v41_dsml.py | 48 | # ============================================================ | COMMENT |
| MEDIUM | c/v41_dsml.py | 50 | # ============================================================ | COMMENT |
| MEDIUM⚡ | c/v41_dsml.py | 108 | # ============================================================ | STRING |
| MEDIUM⚡ | c/v41_dsml.py | 110 | # ============================================================ | STRING |
| MEDIUM | c/v41_dsml.py | 251 | # ============================================================ | COMMENT |
| MEDIUM | c/v41_dsml.py | 253 | # ============================================================ | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 34 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 37 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 65 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 68 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 173 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 175 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 228 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 231 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 258 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 260 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 277 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/diag_harness.py | 279 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 57 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 65 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 81 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 89 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 127 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 135 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 199 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 209 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 349 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/quant_ablation.py | 365 | # -------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/test_olmoe_real.py | 36 | # ── Step 1: Write bootstrap ref with dummy full_ids = prompt_ids ────────── | COMMENT |
| MEDIUM | c/tools/test_olmoe_real.py | 48 | # ── Step 2: Run engine once to capture generated IDs ───────────────────── | COMMENT |
| MEDIUM | c/tools/test_olmoe_real.py | 71 | # ── Step 3: Run engine again against real ref — verify determinism ──────── | COMMENT |
| MEDIUM | c/tools/rans_format.py | 62 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 64 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 199 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 201 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 215 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 217 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 271 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 273 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 320 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 322 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 476 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/rans_format.py | 479 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/expert_atlas/sweep.sh | 8 | # --------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tools/expert_atlas/sweep.sh | 36 | # --------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM | c/tests/test_backend_loader.py | 2483 | # --------------------------------------------------------------------------- | STRING |
| MEDIUM | c/tests/test_backend_loader.py | 2494 | # --------------------------------------------------------------------------- | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | .github/workflows/ci.yml | 165 | # placement without Resizable BAR"), forced here: the harness's results bit for | COMMENT |
| MEDIUM | .github/workflows/ci.yml | 641 | # (observed: 47 min vs a 5-6 min normal run), so the test harness | COMMENT |
| MEDIUM | .github/workflows/ci.yml | 1152 | # Same contract as the Qwen3.6 job runs, from the same harness: a | COMMENT |
| LOW | c/openai_server.py | 5966 | # simply return None for a process (or test double) that only | COMMENT |
| MEDIUM | c/colibri.c | 74 | #include "abl.h" /* per-expert causal-ablation harness — inert unless g_abl.mode set ( | COMMENT |
| MEDIUM | c/tools/efficiency.py | 30 | # Reuse the validated regexes from the existing A/B benchmark harness so the | COMMENT |
| MEDIUM | c/tools/run_f8_bench.sh | 15 | # 8533 MT/s), this harness's primary host. It must be an explicit number | COMMENT |
| MEDIUM | c/tools/fetch_deepgemm.sh | 21 | DEEPGEMM_PIN_DEFAULT=39fb4447a062b418fd08ce17cd308adb28559417 # branch sm120-full, "Add SM120 heuristic calibration ha | CODE |
| MEDIUM | c/tools/bench_fp4_matmul.sh | 6 | # bench_fp4_matmul.c harness against each, verifies the two arms agree | COMMENT |
| MEDIUM | c/tests/vulkan_engines.sh | 85 | # then the expert batch and the weight pool in the same harness, and the routed-expert | COMMENT |
| MEDIUM | c/tests/vulkan_engines.sh | 88 | make tests/test_vk_tier tests/test_glm53_vk_f32 VK=1 # every shader too: the harness's expert batch needs them | CODE |
| MEDIUM | c/tests/vulkan_engines.sh | 101 | # - the harness's results (every format, the tiled GEMMs, the expert batch) bit for bit | COMMENT |
| MEDIUM | c/tests/vulkan_engines.sh | 1780 | # the device and the CPU by measured times); the MTP harness, whose prompt-cache and pin | COMMENT |
| MEDIUM | c/tests/vulkan_engines.sh | 1908 | # The same under ASan and UBSan: every run of the harness's quick subset (depth 3 | COMMENT |
| MEDIUM | c/tests/vulkan_engines.sh | 2202 | # the CPU, serve sessions, and its pin-branch harness with the chain on. | COMMENT |
| MEDIUM | c/tests/vulkan_partial_glm.sh | 20 | # image and its pin-branch harness; | COMMENT |
| MEDIUM | c/tests/vulkan_partial_glm.sh | 388 | # harness compares two engine processes, so the tier is off (which experts a device | COMMENT |
| MEDIUM | c/tests/test_e8x4g64_mint_load.py | 73 | # The Makefile builds this harness with OpenMP on macOS via Homebrew's | COMMENT |
| MEDIUM | c/tests/test_e8x4g64_mint_load.py | 193 | # 5-tuples for the C harness: (name, O, I, wantfmt, wantgs) | COMMENT |
| MEDIUM | c/tests/vulkan_partial_qwen38.sh | 17 | # speculative harness's byte-for-byte gate, serve sessions with prefix reuse and | COMMENT |
| MEDIUM | c/tests/vulkan_partial_qwen38.sh | 268 | # session and the speculative harness's quick subset with 2 of 4 layers on the device. | COMMENT |
| MEDIUM | c/tests/qwen38_mtp_harness.py | 42 | # Q38_MTP_DRAFTS=1: this harness gates the one-draft verify (S=2) and counts its loop; | COMMENT |
| MEDIUM⚡ | c/tests/test_backend_loader.py | 2078 | # Production reused the harness's runtime and released only its own | STRING |
| MEDIUM⚡ | c/tests/test_backend_loader.py | 2079 | # reference, so the harness-owned module survives. | STRING |
| MEDIUM | c/tests/test_backend_loader.py | 868 | # encoding="utf-8": the harness converts every path with | STRING |
| MEDIUM | c/tests/test_backend_loader.py | 1438 | # Production released only its own reference; the harness still holds one. | STRING |
| MEDIUM | c/tests/test_backend_loader.py | 2942 | # The harness still reaches the normal init afterwards: loading on | STRING |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 79 | # well. Without the define the harness died on the #error, so this test | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 127 | # colibri.c is a big translation unit; build the harness once for the class | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 161 | # without either. This harness only exercises the C loader's | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 225 | # Sanity-check the real tool's own selection/stamp BEFORE asking the C harness | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 227 | # the harness anyway would only muddy which side broke. | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 239 | # the fp8-repack path this harness's qt_from_disk-specific decode | COMMENT |
| MEDIUM | c/tests/test_fp8_e2e_repack_load.py | 241 | # st_read_f32 fallback -- out of this C harness's scope, and out of | COMMENT |
| MEDIUM | c/tests/test_python_discovery.py | 20 | # shrinks: rename a harness out of the glob and drop it from here. | COMMENT |
| MEDIUM | c/tests/vulkan_dense_only_glm.sh | 9 | # pin-branch harness, staged uploads, the RAM the experts' cache gets back, and the | COMMENT |
| MEDIUM | c/tests/test_dsv41_prefix_serve.py | 95 | # the arm that wants it has to ask. `reuse=False` still wins: the harness | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | c/openai_server.py | 5106 | (and null when it was described as null), and JSON values are written with | STRING |
| HIGH | c/setup_flow.py | 370 | "build-darwin": "xcode-select --install && brew install libomp", | CODE |
| HIGH | c/setup_flow.py | 692 | script = 'cd "$(cygpath -u "$1")" && shift && exec make "$@"' | CODE |
| HIGH | c/setup_flow.py | 896 | out(" the API still works. To build it: cd web && npm ci && npm run build") | CODE |
| HIGH | c/tools/dsv41_ref.py | 543 | hashes = self.ngram.push(compressed if compressed is not None else ids) | CODE |
| HIGH | c/tools/repack_fp8_passthrough.py | 73 | stamp naming fp8-e4m3-b128 (qt_resolve_fmt's `sf==8 && fp8_blk_f32_also` | STRING |
| HIGH | c/tests/check_data_logprob_gaps.py | 20 | cd c && make glm | STRING |
| HIGH⚡ | c/tests/test_backend_loader_header_parity.py | 62 | self.assertEqual(undefined, [], | CODE |
| HIGH⚡ | c/tests/test_backend_loader_header_parity.py | 64 | "in backend_loader.c: %s" % undefined) | CODE |
| HIGH | c/tests/test_inefficiency.py | 199 | _skip_reason() or "CUDA build not present (run: make clean && make colibri CUDA_DLL=1 && make cuda- | CODE |
| HIGH | c/tests/test_missing_tensor_diagnostics.py | 27 | TINY_GLM_CONFIG = json.loads('{"transformers_version":"5.12.1","architectures":null,"output_hidden_states":false,"return | CODE |
| HIGH | c/tests/test_backend_loader.py | 474 | for (i = 0; i < count && i < MAXI; i++) { | CODE |
| HIGH⚡ | c/tests/test_deepseek_v4_dspark_source.py | 63 | replay = self.engine.index("if (retained > 0 && target_batch(", invalidate) | CODE |
| HIGH⚡ | c/tests/test_glm53_metal_source.py | 29 | self.assertIn("w->fmt == 1 || w->fmt == 4", body) | CODE |
| HIGH⚡ | c/tests/test_engine_evidence.py | 176 | c/quant.h:582 "avx512-vnni" (__AVX512VNNI__ && __AVX512BW__) | STRING |
| HIGH⚡ | c/tests/test_engine_evidence.py | 177 | c/quant.h:584 "avx-vnni" (__AVXVNNI__ && __AVX2__) | STRING |
| HIGH⚡ | c/tests/test_engine_evidence.py | 179 | c/quant.h:588 "neon-i8mm" (__ARM_NEON && __ARM_FEATURE_MATMUL_INT8) | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | web/tests/brio_ui.py | 2 | CODE | |
| LOW | web/tests/brio_ui.py | 2 | CODE | |
| LOW | web/tests/brio_ui.py | 2 | CODE | |
| LOW | colibri/__init__.py | 3 | CODE | |
| LOW | c/experiment_manifest.py | 4 | CODE | |
| LOW | c/autotune.py | 9 | CODE | |
| LOW | c/tools/gen_unicode.py | 9 | CODE | |
| LOW | c/tools/make_qwen38_vision_tiny.py | 21 | CODE | |
| LOW | c/tools/efficiency.py | 22 | CODE | |
| LOW | c/tools/k3_ref.py | 22 | CODE | |
| LOW | c/tools/qpack_http_install.py | 4 | CODE | |
| LOW | c/tools/diag_harness.py | 30 | CODE | |
| LOW | c/tools/make_dsv41_tiny.py | 21 | CODE | |
| LOW | c/tools/convert_olmoe_merged.py | 29 | CODE | |
| LOW | c/tools/convert_qwen36.py | 51 | CODE | |
| LOW | c/tools/convert_qwen36.py | 51 | CODE | |
| LOW | c/tools/make_glm53_multimodal_tiny.py | 22 | CODE | |
| LOW | c/tools/residency_sim.py | 19 | CODE | |
| LOW | c/tools/dsv41_ref.py | 26 | CODE | |
| LOW | c/tools/prepare_dsv41.py | 24 | CODE | |
| LOW | c/tools/make_deepseek_v4_oracle.py | 18 | CODE | |
| LOW | c/tools/make_glm53_tokenizer.py | 20 | CODE | |
| LOW | c/tools/qwenimage_ref.py | 54 | CODE | |
| LOW | c/tools/convert_olmoe.py | 20 | CODE | |
| LOW | c/tools/convert_olmoe.py | 20 | CODE | |
| LOW | c/tools/convert_olmoe.py | 20 | CODE | |
| LOW | c/tools/convert_olmoe.py | 20 | CODE | |
| LOW | c/tools/repair_mtp_int8.py | 35 | CODE | |
| LOW | c/tools/make_glm53_vision_tiny.py | 22 | CODE | |
| LOW | c/tools/qwen38_image.py | 40 | CODE | |
| LOW | c/tools/check_glm53_checkpoint.py | 3 | CODE | |
| LOW | c/tools/make_glm53_tiny.py | 9 | CODE | |
| LOW | c/tools/qpack_install_policy.py | 18 | CODE | |
| LOW | c/tools/qpack_mirror_install.py | 4 | CODE | |
| LOW | c/tools/mw_ur3_raw_adapter.py | 41 | CODE | |
| LOW | c/tools/glm53_image.py | 28 | CODE | |
| LOW | c/tools/make_deepseek_v4_tiny.py | 10 | CODE | |
| LOW | c/tools/dsv41_image.py | 15 | CODE | |
| LOW | c/tools/convert_fp8_to_int4.py | 647 | CODE | |
| LOW | c/tools/convert_fp8_to_int4.py | 648 | CODE | |
| LOW | c/tools/make_qwen38_multimodal_tiny.py | 17 | CODE | |
| LOW | c/tools/gen_qwen38_nfc.py | 4 | CODE | |
| LOW | c/tools/make_glm53_streaming_pair.py | 24 | CODE | |
| LOW | c/tools/make_glm53_streaming_pair.py | 27 | CODE | |
| LOW | c/tools/make_edge_tiny_tokenizer.py | 11 | CODE | |
| LOW | c/tools/expert_atlas/validate.py | 17 | CODE | |
| LOW | c/tests/test_qwen_image_client.py | 7 | CODE | |
| LOW | c/tests/test_qwen_image_client.py | 12 | CODE | |
| LOW | c/tests/test_jev_sdk.py | 211 | CODE | |
| LOW | c/tests/test_mcp_server.py | 15 | CODE | |
| LOW | c/tests/test_deepseek_v4_prefix.py | 14 | CODE | |
| LOW | c/tests/test_deepseek_v4_prefix.py | 21 | CODE | |
| LOW | c/tests/test_qwen_image_cli.py | 26 | CODE | |
| LOW | c/tests/test_family_registry.py | 10 | CODE | |
| LOW | c/tests/test_chat_thinking.py | 14 | CODE | |
| LOW | c/tests/test_chat_thinking.py | 15 | CODE | |
| LOW | c/tests/test_chat_thinking.py | 31 | CODE | |
| LOW | c/tests/test_xdna_package.py | 19 | CODE | |
| LOW | c/tests/test_inefficiency.py | 18 | CODE | |
| LOW | c/tests/test_inefficiency.py | 22 | CODE | |
| 12 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | c/tests/test_qwen36_dashboard.py | 0 | hits and prof follow done on the engine's stdout, so the reply can reach the client before the gateway has read them. wa | STRING |
| HIGH | c/tests/test_glm53_dashboard.py | 0 | hits and prof follow done on the engine's stdout, so the reply can reach the client before the gateway has read them. wa | STRING |
| HIGH | c/tests/test_dsv41_dashboard.py | 0 | hits and prof follow done on the engine's stdout, so the reply can reach the client before the gateway has read them. wa | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | c/openai_server.py | 8167 | CODE | |
| LOW | c/openai_server.py | 5721 | CODE | |
| LOW | c/autotune.py | 413 | CODE | |
| LOW | c/doctor.py | 572 | CODE | |
| LOW | c/resource_plan.py | 2942 | CODE | |
| LOW | c/tools/qpack_http_install.py | 1128 | CODE | |
| LOW | c/tools/qpack_http_install.py | 1146 | CODE | |
| LOW | c/tools/make_qwen38_tiny.py | 323 | CODE | |
| LOW | c/tools/gguf_reader.py | 370 | CODE | |
| LOW | c/tools/make_kimi_k3_ref.py | 144 | CODE | |
| LOW | c/tools/compare_laya.py | 219 | CODE | |
| LOW | c/tools/make_qwen36_tiny.py | 162 | CODE | |
| LOW | c/tests/test_cgroup_memory.py | 73 | CODE | |
| LOW | c/tests/test_check_data_logprob_gaps.py | 78 | CODE | |
| LOW | c/tests/test_mw_ur3_raw_adapter.py | 253 | CODE | |
| LOW | c/tests/test_systemone_api.py | 52 | CODE | |
| LOW | c/tests/test_openai_server.py | 108 | CODE | |
| LOW | c/tests/test_openai_server.py | 129 | CODE | |
| LOW | c/tests/test_openai_server.py | 279 | CODE | |
| LOW | c/tests/test_openai_server.py | 330 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | docs/kimi_metal_gap_analysis.md | 293 | ### Phase 4 Changes (scaffolding only) | COMMENT |
| MEDIUM | c/openai_server.py | 2627 | # preamble makes it hallucinate other frameworks' syntax (e.g. `end_action`). | COMMENT |
| MEDIUM | c/tests/test_registry_engine_agreement.py | 57 | # No context window to cap: the launcher refuses --ctx for an | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | docker/README.md | 67 | ### Step 1: Download the model | COMMENT |
| LOW | docker/README.md | 98 | ### Step 2: Download the Colibrì Dockerfile | COMMENT |
| LOW | docker/README.md | 106 | ### Step 3: Build the Docker image | COMMENT |
| LOW | docker/README.md | 135 | ### Step 4: Start Colibrì | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | c/chat_olmoe.sh | 6 | # Usage: | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | colibri/__init__.py | 5 | __all__ = ["__version__"] | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | c/tools/make_laya_tiny.py | 93 | {"subject": "Duplicate charge", "body": "Please fix this", "from": "user@example.com"} | CODE |