{"id":"pathwaycom-bdh","name":"bdh","homepage":null,"repo_url":"https://github.com/pathwaycom/bdh","category":"ai-ml","subcategories":[],"tags":["ai-ml","llm-architecture","transformers","interpretability","neuroscience-inspired","python","research"],"what_it_does":"Baby Dragon Hatchling (BDH) is an open-source Python repository implementing a biologically inspired neural architecture (per the linked paper) intended to bridge transformer-like performance with neuroscience-motivated, locally interacting neuron dynamics and interpretable sparse activations.","use_cases":["Research and prototyping of alternative LLM/transformer-like architectures","Investigating interpretability via neuron-level sparse/positive activations","Training small-to-medium models for language/translation tasks at reported GPT-2–scale parameter regimes","Experimentation on scaling behavior and “brain-inspired” reasoning dynamics"],"not_for":["Production deployment without additional engineering, testing, and evaluation hardening","Use as a hosted API/service (no evidence of an online endpoint)","Security-sensitive environments without auditing dependencies and training/inference code paths"],"best_when":"You want to study/experiment with model architecture ideas from the referenced paper and run training locally (e.g., GPU environments) rather than integrate via an external API.","avoid_when":"You need a turnkey, API-first developer experience (REST/SDK), managed uptime guarantees, or well-specified auth/rate-limit behavior from a hosted service.","alternatives":["nanoGPT (for transformer training baselines)","Other open transformer implementations (e.g., Hugging Face Transformers models as baselines)","Community ports mentioned in the repository README (different language/stack implementations)"],"af_score":29.0,"security_score":18.0,"reliability_score":20.0,"package_type":"skill","discovery_source":["openclaw"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-29T18:04:38.727743+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":false,"mcp_server_url":null,"has_sdk":false,"sdk_languages":[],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":[],"oauth":false,"scopes":false,"notes":"No hosted interface described; authentication is not applicable to this repository’s local training scripts."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Repository appears to be open-source; compute costs for training/inference depend on user hardware and dataset choices."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":29.0,"security_score":18.0,"reliability_score":20.0,"mcp_server_quality":0.0,"documentation_accuracy":45.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":100.0,"rate_limit_clarity":0.0,"tls_enforcement":0.0,"auth_strength":0.0,"scope_granularity":0.0,"dependency_hygiene":40.0,"secret_handling":60.0,"security_notes":"No hosted network surface is described. Security posture depends on local code execution and the contents of requirements.txt and any data paths. Given only high-level README content, dependency hygiene and secret handling cannot be fully verified; agents should audit dependencies, review training scripts for logging of secrets, and run static/dynamic analysis before use.","uptime_documented":0.0,"version_stability":30.0,"breaking_changes_history":30.0,"error_recovery":20.0,"idempotency_support":"false","idempotency_notes":null,"pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["No evidence of a programmatic API or MCP server; agents must interact by cloning/running code locally.","README provides only high-level install/train commands; detailed CLI parameters, config formats, and expected artifacts are not provided in the supplied content.","ML training/inference has non-determinism and environment sensitivity (GPU/seed/config), which can reduce “agent reliability” without robust experiment management."]}}