hypermind 1.4.6__tar.gz → 1.4.7__tar.gz

{hypermind-1.4.6 → hypermind-1.4.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hypermind
-Version: 1.4.6
+Version: 1.4.7
 Summary: Federated agent intelligence — reference SDK for the HyperMind SCITT contested-claims profile
 Project-URL: Homepage, https://github.com/ZySec-AI/hypermind
 Project-URL: Documentation, https://github.com/ZySec-AI/hypermind/tree/main/docs
@@ -137,7 +137,7 @@ Native adapters: **LangChain** · **CrewAI** · **AutoGen** · **LlamaIndex** ·
 
 ---
 
-## MCP tools (34 tools)
+## MCP tools (38 tools)
 
 Every tool below is available when you run `hmctl mcp install`. Read-only by default;
 `hm_publish_claim` requires `--write`.
@@ -151,7 +151,7 @@ Every tool below is available when you run `hmctl mcp install`. Read-only by def
 | `hm_recall` | Memory search — "what does the swarm believe about X?" with bi-temporal time-travel |
 | `hm_related` | Graph traversal: find connected claims via CITES / CORROBORATES links |
 | `hm_learn` | Trigger focused OrgSim research on a topic; publishes signed findings to the ledger |
-| `hm_publish_claim` | Publish a signed, attributed claim to the ledger **(requires `--write`)** |
+| `hm_publish_claim` | Publish a signed, attributed claim to the ledger with optional `ttl_days` for knowledge decay **(requires `--write`)** |
 
 ### Ask the network
 
@@ -169,7 +169,8 @@ Every tool below is available when you run `hmctl mcp install`. Read-only by def
 | Tool | What it does |
 |---|---|
 | `hm_ingest` | Ingest `.md` / `.txt` files or directories into the swarm as signed claims |
-| `hm_dataset_export` | Export the verified corpus as SFT / DPO / reward-model fine-tuning JSONL (supports `--push-to-hub`) |
+| `hm_dataset_export` | Export the verified corpus as JSONL / Parquet / HuggingFace dataset with SHA-256 digest, Merkle proofs, and `--exclude-contested` filter (supports `--push-to-hub`) |
+| `hm_stale` | List claims past their TTL with effective confidence after knowledge-decay scoring |
 
 Domain packs: `hypermind.domains.medical`, `.defense`, `.legal` — pre-configured populations + prompts for verticals.
 
@@ -211,6 +212,23 @@ Domain packs: `hypermind.domains.medical`, `.defense`, `.legal` — pre-configur
 | `hm_org_roster` | Describe a template's structure — units, seats, concrete roster |
 | `hm_run_org` | Server-free alias of `hm_org_roster` |
 
+### Autonomous learning engine
+
+| Tool | What it does |
+|---|---|
+| `hm_loop` | Start / stop / inspect an always-on autonomous learning loop per topic — runs question → research → publish → contradiction-scan cycles on a schedule |
+| `hm_next_question` | Self-directed question generation: picks the highest-value open question via `uncertainty`, `contradiction`, or `frontier` strategy over the HypothesisBook |
+| `hm_scan_contradictions` | Lexical contradiction detection (Jaccard + negation-asymmetry) across the knowledge base; optional `auto_dispute=true` to raise disputes automatically |
+| `hm_calibration` | Per-agent Brier / ECE calibration report — shows how well each agent's stated confidence predicts outcomes, with a `calibration_weight` applied to future reputation updates |
+| `hm_federate` | Connect to a peer namespace over TCP (`connect` / `disconnect` / `pull` / `status`); syncs claims via anti-entropy Bloom reconciliation; foreign reputation is display-only |
+
+### Observability
+
+| Tool | What it does |
+|---|---|
+| `hm_status` | Feature-by-feature health check — OK / degraded / unconfigured with actionable `fix` instructions for each |
+| `hm_stats` | Knowledge-base snapshot: claim count, confidence distribution, agents by tier, topics by expert count, evolution timeline |
+
 ### Meta
 
 | Tool | What it does |
@@ -244,7 +262,7 @@ Built on **COSE** (RFC 9052), **Merkle transparency** (RFC 6962), **FROST thresh
 |---|---|---|
 | **Deep research engine** | `hypermind.orgs` | PI decomposition + self-refine, 12 reasoning strategies, hypothesis lifecycle, architecture-native metrics, streaming learn-large/serve-small query, full-run replay → emits a **trust-scored verified corpus** |
 | **Eval suite** | `hypermind.eval` | Proper scoring rules (Brier, ECE, EIG/Q, skill score, Murphy-Winkler decomposition), self-contained HTML reports, `hmctl eval` CLI |
-| **Adoption layer** | `hypermind.integrations` | One-object facade, OpenAI-compatible endpoint, 34-tool MCP server, framework adapters (LangChain / CrewAI / AutoGen / LlamaIndex / OpenAI Agents), knowledge ingestion, fine-tuning dataset export, domain packs |
+| **Adoption layer** | `hypermind.integrations` | One-object facade, OpenAI-compatible endpoint, 38-tool MCP server, framework adapters (LangChain / CrewAI / AutoGen / LlamaIndex / OpenAI Agents), knowledge ingestion, fine-tuning dataset export, autonomous learning engine (`hm_loop`, `hm_next_question`, `hm_scan_contradictions`, `hm_calibration`, `hm_federate`), domain packs |
 
 **The learn-large / serve-small loop:** feed a small model high-trust, contested-and-survived knowledge
 from the verified corpus and it outperforms a larger model trained on raw web text.

{hypermind-1.4.6 → hypermind-1.4.7}/README.md

@@ -65,7 +65,7 @@ Native adapters: **LangChain** · **CrewAI** · **AutoGen** · **LlamaIndex** ·
 
 ---
 
-## MCP tools (34 tools)
+## MCP tools (38 tools)
 
 Every tool below is available when you run `hmctl mcp install`. Read-only by default;
 `hm_publish_claim` requires `--write`.
@@ -79,7 +79,7 @@ Every tool below is available when you run `hmctl mcp install`. Read-only by def
 | `hm_recall` | Memory search — "what does the swarm believe about X?" with bi-temporal time-travel |
 | `hm_related` | Graph traversal: find connected claims via CITES / CORROBORATES links |
 | `hm_learn` | Trigger focused OrgSim research on a topic; publishes signed findings to the ledger |
-| `hm_publish_claim` | Publish a signed, attributed claim to the ledger **(requires `--write`)** |
+| `hm_publish_claim` | Publish a signed, attributed claim to the ledger with optional `ttl_days` for knowledge decay **(requires `--write`)** |
 
 ### Ask the network
 
@@ -97,7 +97,8 @@ Every tool below is available when you run `hmctl mcp install`. Read-only by def
 | Tool | What it does |
 |---|---|
 | `hm_ingest` | Ingest `.md` / `.txt` files or directories into the swarm as signed claims |
-| `hm_dataset_export` | Export the verified corpus as SFT / DPO / reward-model fine-tuning JSONL (supports `--push-to-hub`) |
+| `hm_dataset_export` | Export the verified corpus as JSONL / Parquet / HuggingFace dataset with SHA-256 digest, Merkle proofs, and `--exclude-contested` filter (supports `--push-to-hub`) |
+| `hm_stale` | List claims past their TTL with effective confidence after knowledge-decay scoring |
 
 Domain packs: `hypermind.domains.medical`, `.defense`, `.legal` — pre-configured populations + prompts for verticals.
 
@@ -139,6 +140,23 @@ Domain packs: `hypermind.domains.medical`, `.defense`, `.legal` — pre-configur
 | `hm_org_roster` | Describe a template's structure — units, seats, concrete roster |
 | `hm_run_org` | Server-free alias of `hm_org_roster` |
 
+### Autonomous learning engine
+
+| Tool | What it does |
+|---|---|
+| `hm_loop` | Start / stop / inspect an always-on autonomous learning loop per topic — runs question → research → publish → contradiction-scan cycles on a schedule |
+| `hm_next_question` | Self-directed question generation: picks the highest-value open question via `uncertainty`, `contradiction`, or `frontier` strategy over the HypothesisBook |
+| `hm_scan_contradictions` | Lexical contradiction detection (Jaccard + negation-asymmetry) across the knowledge base; optional `auto_dispute=true` to raise disputes automatically |
+| `hm_calibration` | Per-agent Brier / ECE calibration report — shows how well each agent's stated confidence predicts outcomes, with a `calibration_weight` applied to future reputation updates |
+| `hm_federate` | Connect to a peer namespace over TCP (`connect` / `disconnect` / `pull` / `status`); syncs claims via anti-entropy Bloom reconciliation; foreign reputation is display-only |
+
+### Observability
+
+| Tool | What it does |
+|---|---|
+| `hm_status` | Feature-by-feature health check — OK / degraded / unconfigured with actionable `fix` instructions for each |
+| `hm_stats` | Knowledge-base snapshot: claim count, confidence distribution, agents by tier, topics by expert count, evolution timeline |
+
 ### Meta
 
 | Tool | What it does |
@@ -172,7 +190,7 @@ Built on **COSE** (RFC 9052), **Merkle transparency** (RFC 6962), **FROST thresh
 |---|---|---|
 | **Deep research engine** | `hypermind.orgs` | PI decomposition + self-refine, 12 reasoning strategies, hypothesis lifecycle, architecture-native metrics, streaming learn-large/serve-small query, full-run replay → emits a **trust-scored verified corpus** |
 | **Eval suite** | `hypermind.eval` | Proper scoring rules (Brier, ECE, EIG/Q, skill score, Murphy-Winkler decomposition), self-contained HTML reports, `hmctl eval` CLI |
-| **Adoption layer** | `hypermind.integrations` | One-object facade, OpenAI-compatible endpoint, 34-tool MCP server, framework adapters (LangChain / CrewAI / AutoGen / LlamaIndex / OpenAI Agents), knowledge ingestion, fine-tuning dataset export, domain packs |
+| **Adoption layer** | `hypermind.integrations` | One-object facade, OpenAI-compatible endpoint, 38-tool MCP server, framework adapters (LangChain / CrewAI / AutoGen / LlamaIndex / OpenAI Agents), knowledge ingestion, fine-tuning dataset export, autonomous learning engine (`hm_loop`, `hm_next_question`, `hm_scan_contradictions`, `hm_calibration`, `hm_federate`), domain packs |
 
 **The learn-large / serve-small loop:** feed a small model high-trust, contested-and-survived knowledge
 from the verified corpus and it outperforms a larger model trained on raw web text.

{hypermind-1.4.6 → hypermind-1.4.7}/docs/feature-matrix.md

@@ -400,6 +400,23 @@ Swarm reasoning tools, belief weighting, and research persistence.
 | Research job persistence (`~/.hypermind/jobs/<job_id>.json`) | Stable | 1.4.0 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_research_tool.py` | — | — |
 | `hypermind_llm()` CrewAI adapter | Stable | 1.4.0 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_crewai.py` | — | — |
 
+### Autonomous learning engine (v1.4.6)
+
+| Capability | Status | Since | Until | Spec § | Test file | Example | RuleId(s) |
+|---|---|---|---|---|---|---|---|
+| `hm_loop` MCP tool — always-on autonomous learning scheduler; asyncio task + `~/.hypermind/loops/` persistence | Stable | 1.4.6 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_next_question` MCP tool — self-directed question generation (uncertainty / contradiction / frontier strategies) | Stable | 1.4.6 | — | [§20-swarm-intelligence](spec/20-swarm-intelligence.md) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_scan_contradictions` MCP tool — Jaccard + negation-asymmetry contradiction detection; `auto_dispute` flag | Stable | 1.4.6 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_calibration` MCP tool — per-agent Brier / ECE calibration report; `calibration_weight` in `ReputationKernel` | Stable | 1.4.6 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_stale` MCP tool — claims past TTL with freshness-decay effective confidence | Stable | 1.4.6 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_federate` MCP tool — TCP peer connect/disconnect/pull/status; Bloom anti-entropy; foreign reputation display-only | Stable | 1.4.6 | — | [§01-transport](spec/01-transport.md) | `tests/integrations/test_autonomous.py` | — | — |
+| Knowledge decay — `ttl_days` on `hm_publish_claim`; `freshness_factor()` linear decay 1.0→0.5 after half-TTL | Stable | 1.4.6 | — | [§04-knowledge-model](spec/04-knowledge-model.md) | `tests/integrations/test_autonomous.py` | — | — |
+| `hm_dataset_export` extended — SHA-256 digest, Merkle proof, Parquet / HuggingFace output, `exclude_contested` | Stable | 1.4.6 | — | [§09-agent-sdk](spec/09-agent-sdk.md#mcp-bridge) | `tests/integrations/test_mcp_tools.py` | — | — |
+| `hm_status` MCP tool — feature health check with actionable `fix` per degraded feature | Stable | 1.4.5 | — | [§13-observability](spec/13-observability.md) | `tests/integrations/test_mcp_tools.py` | — | — |
+| `hm_stats` MCP tool — knowledge-base snapshot, agents by tier, topics, evolution timeline | Stable | 1.4.5 | — | [§13-observability](spec/13-observability.md) | `tests/integrations/test_mcp_tools.py` | — | — |
+| `hmctl loop start/stop/status` CLI | Stable | 1.4.6 | — | — | — | — | — |
+| `hmctl explore <topic>` CLI — autonomous research loop with Rich progress display | Stable | 1.4.6 | — | — | — | — | — |
+
 ## Deferred to external (ecosystem gates)
 
 Code or hooks are shipped; the remaining work is operator, foundation, or

{hypermind-1.4.6 → hypermind-1.4.7}/docs/integrations.md

@@ -199,37 +199,91 @@ hmctl mcp serve               # run the MCP server directly (for other hosts)
 best available model, and writes `~/.hypermind/ledger.db` as the persistent
 ledger — no API key, no internet required.
 
-### MCP tool reference
+### MCP tool reference (38 tools)
 
-| Tool | Category | Description |
-|---|---|---|
-| `hm_version` | Info | SDK version and feature flags |
-| `hm_tools` | Info | List all available MCP tools with descriptions |
-| `hm_ask` | Network | Ask a question; returns a reputation-weighted swarm answer |
-| `hm_retrieve` | Retrieval | Trust-scored corpus retrieval (like a vector search, but with provenance) |
-| `hm_recall` | Retrieval | Semantic search over in-memory + Relata claims |
-| `hm_justify` | Retrieval | Full provenance for a claim: author, confidence, dispute history |
-| `hm_related` | Retrieval | Find claims related to a query (requires `RELATA_URL`) |
-| `hm_learn` | Write | Publish a finding into the swarm; requires `--write` flag |
-| `hm_publish_claim` | Write | Low-level claim publish with explicit confidence; requires `--write` flag |
-| `hm_reputation` | Trust | Agent's earned reputation on a topic (Beta posterior → float) |
-| `hm_roster` | Trust | List agents in the swarm with their roles and reputation |
-| `hm_agent` | Trust | Details for a specific agent by `kid` hex |
-| `hm_topic_experts` | Trust | Top-N agents by reputation on a topic |
-| `hm_disputes` | Disputes | Active and resolved disputes; FSM state per claim |
-| `hm_research_start` | Research | Kick off an async OrgSim research run; returns a run ID |
-| `hm_research_status` | Research | Poll a research run by ID |
-| `hm_research_result` | Research | Retrieve the completed research report |
-| `hm_population_list` | Population | List saved population checkpoints (`~/.hypermind/populations/`) |
-| `hm_population_get` | Population | Load a checkpoint — personas, reputation state, claim count |
-| `hm_list_templates` | Orgs | Available org templates (research_lab, debate_club, …) |
-| `hm_org_roster` | Orgs | Agents in a running OrgSim |
-| `hm_run_org` | Orgs | Run a full OrgSim and return the report synchronously |
-
-**Research tools always available.** `hm_learn`, `hm_research_start/status/result` are
-registered on all installs — they need an LLM key to do useful work but won't silently
-fail on a read-only install. Only `hm_publish_claim` (raw ledger mutation) requires
-`hmctl mcp install --write`. The default install is safe to hand to an AI assistant.
+**Observability**
+
+| Tool | Description |
+|---|---|
+| `hm_version` | SDK version, wire version, read/write mode, registered tool list |
+| `hm_status` | Feature-by-feature health check — OK / degraded / unconfigured with actionable `fix` instructions |
+| `hm_stats` | Knowledge-base snapshot: claim count, confidence distribution, agents by tier, topics, evolution timeline |
+| `hm_tools` | List all available MCP tools with descriptions |
+
+**Knowledge & retrieval**
+
+| Tool | Description |
+|---|---|
+| `hm_retrieve` | Trust-scored BM25 hybrid search over the signed ledger |
+| `hm_recall` | Semantic search over in-memory + Relata claims |
+| `hm_justify` | Full provenance chain for a claim: author, confidence, dispute history, citations |
+| `hm_related` | Graph traversal: connected claims via CITES / CORROBORATES links |
+| `hm_learn` | Publish a finding into the swarm *(requires `--write`)* |
+| `hm_publish_claim` | Low-level signed claim publish with explicit confidence and optional `ttl_days` *(requires `--write`)* |
+| `hm_ingest` | Ingest `.md` / `.txt` files as signed claims |
+| `hm_dataset_export` | Export corpus as JSONL / Parquet / HF dataset with SHA-256 digest, Merkle proofs, `--exclude-contested` |
+| `hm_stale` | List claims past their TTL with effective confidence after knowledge-decay scoring |
+
+**Ask the network**
+
+| Tool | Description |
+|---|---|
+| `hm_ask` | Reputation-weighted panel deliberation — one answer with provenance |
+| `hm_simulate` | Counterfactual what-if: confidence delta between original and counterfactual |
+| `hm_debate` | Adversarial pro/con panel for N rounds |
+| `hm_cross_check` | Multi-source cross-validation: support / refute / uncertain |
+| `hm_hypothesis` | Register a falsifiable hypothesis |
+| `hm_goal` | Goal-directed swarm planning |
+
+**Autonomous learning engine**
+
+| Tool | Description |
+|---|---|
+| `hm_loop` | Start / stop / inspect an always-on autonomous learning loop per topic — runs question → research → publish → contradiction-scan on a schedule |
+| `hm_next_question` | Self-directed question generation: `uncertainty`, `contradiction`, or `frontier` strategy over the HypothesisBook |
+| `hm_scan_contradictions` | Lexical contradiction detection (Jaccard + negation-asymmetry); optional `auto_dispute=true` |
+| `hm_calibration` | Per-agent Brier / ECE calibration report with `calibration_weight` for future reputation updates |
+| `hm_federate` | Connect to peer namespaces over TCP; sync via Bloom anti-entropy; foreign reputation display-only |
+
+**Trust & reputation**
+
+| Tool | Description |
+|---|---|
+| `hm_reputation` | Agent's earned reputation on a topic (Beta posterior → float) |
+| `hm_roster` | List all agents with role, rooms, reputation, ramp-complete flag |
+| `hm_agent` | Details for a specific agent by `kid` hex |
+| `hm_topic_experts` | Top-N agents by reputation on a topic |
+| `hm_disputes` | Active and resolved disputes; FSM state per claim |
+| `hm_audit` | Tamper-evident audit log summary |
+
+**Research engine**
+
+| Tool | Description |
+|---|---|
+| `hm_research_start` | Kick off an async OrgSim research run; returns a run ID |
+| `hm_research_status` | Poll a research run by ID |
+| `hm_research_result` | Retrieve the completed research report |
+
+**Persistent populations**
+
+| Tool | Description |
+|---|---|
+| `hm_population_create` | Scaffold a named population from a template |
+| `hm_population_load` | Rehydrate a saved population with accumulated reputation |
+| `hm_population_checkpoint` | Snapshot current swarm reputation / claim counts |
+| `hm_population_list` | List all saved populations |
+| `hm_population_get` | Detail a population's personas, roles, topics, claim counts |
+
+**Evaluation & org templates**
+
+| Tool | Description |
+|---|---|
+| `hm_eval` | Proper-score calibration suite (Brier, ECE, log-score, WoC-Δ, EIG/Q) |
+| `hm_list_templates` | Available org templates (research_lab, debate_club, …) |
+| `hm_org_roster` | Describe a template's structure — units, seats, roster |
+| `hm_run_org` | Run a full OrgSim and return the report synchronously |
+
+**Write access note:** Only `hm_publish_claim`, `hm_learn`, and `hm_ingest` mutate the ledger — they require `hmctl mcp install --write`. All other tools are read-only and safe to hand to any AI assistant.
 
 ### Environment variables for the MCP server
 

{hypermind-1.4.6 → hypermind-1.4.7}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "hypermind"
-version = "1.4.6"
+version = "1.4.7"
 description = "Federated agent intelligence — reference SDK for the HyperMind SCITT contested-claims profile"
 readme = "README.md"
 requires-python = ">=3.11,<3.14"

{hypermind-1.4.6 → hypermind-1.4.7}/src/hmctl/cli.py

@@ -464,9 +464,9 @@ def status_cmd() -> None:
     """Show what's configured, what's degraded, and how to fix it."""
     import os as _os
 
+    from rich import box
     from rich.console import Console
     from rich.table import Table
-    from rich import box
 
     console = Console()
 
@@ -514,19 +514,39 @@ def status_cmd() -> None:
 
     rows = [
         # (feature, ok, detail, fix)
-        ("env_file", has_env_file,
-         "~/.hypermind/env loaded" if has_env_file else "~/.hypermind/env not found — API keys must be exported every session",
-         "touch ~/.hypermind/env  # one KEY=VALUE per line"),
+        (
+            "env_file",
+            has_env_file,
+            "~/.hypermind/env loaded"
+            if has_env_file
+            else "~/.hypermind/env not found — API keys must be exported every session",
+            "touch ~/.hypermind/env  # one KEY=VALUE per line",
+        ),
         ("llm", llm_ok, llm_note, llm_fix),
-        ("relata", bool(relata_url),
-         relata_url if relata_url else "No durable ledger — claims are in-process only, lost on restart; hm_related falls back to keyword scan",
-         "1. bash deploy/relata-serve.sh\n2. Add RELATA_URL=http://localhost:5432 to ~/.hypermind/env\n3. Restart hmctl mcp serve"),
-        ("otel", bool(otel_ep),
-         otel_ep if otel_ep else "No OTLP endpoint — audit trail is local-only, not shipped to your observability stack",
-         "Add OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 to ~/.hypermind/env\n(Jaeger / Grafana Tempo / Honeycomb / Datadog all accept OTLP)"),
-        ("sqlite_cache", bool(db_path),
-         db_path if db_path else "No local DB — in-process only; all claims reset on server restart",
-         "Add HYPERMIND_DB=~/.hypermind/cache.db to ~/.hypermind/env"),
+        (
+            "relata",
+            bool(relata_url),
+            relata_url
+            if relata_url
+            else "No durable ledger — claims are in-process only, lost on restart; hm_related falls back to keyword scan",
+            "1. bash deploy/relata-serve.sh\n2. Add RELATA_URL=http://localhost:5432 to ~/.hypermind/env\n3. Restart hmctl mcp serve",
+        ),
+        (
+            "otel",
+            bool(otel_ep),
+            otel_ep
+            if otel_ep
+            else "No OTLP endpoint — audit trail is local-only, not shipped to your observability stack",
+            "Add OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 to ~/.hypermind/env\n(Jaeger / Grafana Tempo / Honeycomb / Datadog all accept OTLP)",
+        ),
+        (
+            "sqlite_cache",
+            bool(db_path),
+            db_path
+            if db_path
+            else "No local DB — in-process only; all claims reset on server restart",
+            "Add HYPERMIND_DB=~/.hypermind/cache.db to ~/.hypermind/env",
+        ),
     ]
 
     table = Table(title=f"hmctl status — HyperMind {__version__}", box=box.ROUNDED, show_lines=True)
@@ -1332,13 +1352,20 @@ def loop_group() -> None:
 
 @loop_group.command("start")
 @click.argument("topic")
-@click.option("--every", "interval_hours", default=4.0, show_default=True, help="Hours between cycles.")
+@click.option(
+    "--every", "interval_hours", default=4.0, show_default=True, help="Hours between cycles."
+)
 @click.option("--rounds", default=3, show_default=True, help="OrgSim research rounds per cycle.")
 def loop_start(topic: str, interval_hours: float, rounds: int) -> None:
     """Start an autonomous research loop on TOPIC."""
-    from hypermind.integrations.loop import start_loop, save_loop_state
+    from hypermind.integrations.loop import save_loop_state
 
-    state = {"topic": topic, "interval_hours": interval_hours, "rounds": rounds, "status": "scheduled"}
+    state = {
+        "topic": topic,
+        "interval_hours": interval_hours,
+        "rounds": rounds,
+        "status": "scheduled",
+    }
     save_loop_state(state)
     click.echo(f"Loop scheduled for topic '{topic}' every {interval_hours}h.")
     click.echo(f"  State: ~/.hypermind/loops/{topic}.json")
@@ -1363,9 +1390,10 @@ def loop_stop(topic: str) -> None:
 @click.argument("topic", required=False, default=None)
 def loop_status_cmd(topic: str | None) -> None:
     """Show status of all loops (or a specific TOPIC)."""
+    from rich import box
     from rich.console import Console
     from rich.table import Table
-    from rich import box
+
     from hypermind.integrations.loop import loop_status
 
     loops = loop_status(topic)
@@ -1380,15 +1408,17 @@ def loop_status_cmd(topic: str | None) -> None:
     table.add_column("Cycles")
     table.add_column("Last run")
     table.add_column("Next run")
-    for l in loops:
-        status_str = "[green]running[/green]" if l.get("running") else l.get("status", "unknown")
+    for loop in loops:
+        status_str = (
+            "[green]running[/green]" if loop.get("running") else loop.get("status", "unknown")
+        )
         table.add_row(
-            l.get("topic", ""),
+            loop.get("topic", ""),
             status_str,
-            f"{l.get('interval_hours', '?')}h",
-            str(l.get("cycles_completed", 0)),
-            l.get("last_run", "—"),
-            l.get("next_run", "—"),
+            f"{loop.get('interval_hours', '?')}h",
+            str(loop.get("cycles_completed", 0)),
+            loop.get("last_run", "—"),
+            loop.get("next_run", "—"),
         )
     console.print(table)
 
@@ -1397,7 +1427,12 @@ def loop_status_cmd(topic: str | None) -> None:
 @click.argument("topic")
 @click.option("--cycles", default=3, show_default=True, help="Number of research cycles to run.")
 @click.option("--rounds", default=2, show_default=True, help="OrgSim rounds per cycle.")
-@click.option("--strategy", type=click.Choice(["uncertainty", "contradiction", "frontier"]), default="uncertainty", show_default=True)
+@click.option(
+    "--strategy",
+    type=click.Choice(["uncertainty", "contradiction", "frontier"]),
+    default="uncertainty",
+    show_default=True,
+)
 @click.option("--swarm", default=3, show_default=True)
 def explore_cmd(topic: str, cycles: int, rounds: int, strategy: str, swarm: int) -> None:
     """Deep-explore a topic: hypothesis → research → validate → repeat.
@@ -1406,6 +1441,7 @@ def explore_cmd(topic: str, cycles: int, rounds: int, strategy: str, swarm: int)
     scanning for contradictions, and building up the knowledge base.
     """
     import asyncio
+
     from rich.console import Console
     from rich.progress import Progress, SpinnerColumn, TextColumn
 
@@ -1419,26 +1455,34 @@ def explore_cmd(topic: str, cycles: int, rounds: int, strategy: str, swarm: int)
         client = HyperMindClient.sandbox(swarm=swarm, topic=topic, responder=responder)
         from hypermind.integrations.loop import _run_cycle
 
-        console.print(f"\n[bold]Exploring:[/bold] {topic}  ({cycles} cycles, {rounds} rounds each)\n")
+        console.print(
+            f"\n[bold]Exploring:[/bold] {topic}  ({cycles} cycles, {rounds} rounds each)\n"
+        )
         total_published = 0
         total_contradictions = 0
 
-        with Progress(SpinnerColumn(), TextColumn("[progress.description]{task.description}"), console=console) as prog:
+        with Progress(
+            SpinnerColumn(), TextColumn("[progress.description]{task.description}"), console=console
+        ) as prog:
             for i in range(cycles):
-                task = prog.add_task(f"Cycle {i+1}/{cycles}", total=None)
+                task = prog.add_task(f"Cycle {i + 1}/{cycles}", total=None)
                 result = await _run_cycle(topic, client, responder, rounds)
                 prog.remove_task(task)
                 total_published += result.get("published", 0)
                 total_contradictions += result.get("contradictions_found", 0)
                 console.print(
-                    f"  [cyan]cycle {i+1}[/cyan]  "
+                    f"  [cyan]cycle {i + 1}[/cyan]  "
                     f"question: {result.get('question', '')[:70]}…  "
                     f"published: {result.get('published', 0)}  "
                     f"contradictions: {result.get('contradictions_found', 0)}"
                 )
 
-        console.print(f"\n[green]Done.[/green] {total_published} claims published, {total_contradictions} contradictions detected.")
-        console.print(f"Knowledge base now has {len(client._claims)} total claims on topic '{topic}'.")
+        console.print(
+            f"\n[green]Done.[/green] {total_published} claims published, {total_contradictions} contradictions detected."
+        )
+        console.print(
+            f"Knowledge base now has {len(client._claims)} total claims on topic '{topic}'."
+        )
         await client.close()
 
     asyncio.run(_run())
@@ -1513,9 +1557,10 @@ def mcp_serve_cmd(
         # Welcome banner → stderr only (stdout is the MCP stdio channel; anything
         # printed there corrupts the JSON-RPC stream). Shows up in the host's MCP
         # server log so you can confirm which build is live.
+        import os as _os
+
         from hypermind import __version__
 
-        import os as _os
         ident = sign_as.kid.hex()[:16] + "…" if sign_as is not None else "anonymous"
         llm = getattr(responder, "model", None) if responder is not None else None
         base_url = _os.environ.get("OPENAI_BASE_URL", "")
@@ -1527,7 +1572,7 @@ def mcp_serve_cmd(
         lines = [
             f" ╦ ╦╦ ╦╔═╗╔═╗╦═╗╔╦╗╦╔╗╔╔╦╗   MCP v{__version__} — ready",
             f" ╠═╣╚╦╝╠═╝║╣ ╠╦╝║║║║║║║ ║║   {len(server.tools)} tools · {'read-write' if writable else 'read-only'} · {ident}",
-            f" ╩ ╩ ╩ ╩  ╚═╝╩╚═╩ ╩╩╝╚╝═╩╝",
+            " ╩ ╩ ╩ ╩  ╚═╝╩╚═╩ ╩╩╝╚╝═╩╝",
             "",
             f"  llm     {'✓ ' + llm + ('  (' + base_url + ')' if base_url else '') if llm else '✗ off  →  add OPENAI_API_KEY + HYPERMIND_LLM_MODEL to ~/.hypermind/env'}",
             f"  config  {'✓ ~/.hypermind/env' if has_env_file else '✗ not found  →  touch ~/.hypermind/env  (KEY=VALUE, one per line)'}",
@@ -1763,15 +1808,26 @@ def dataset_group() -> None:
 @click.option("--out", default=None, help="Output file path.")
 @click.option("--out-dir", default=None, help="Output directory (writes all three formats).")
 @click.option("--all", "export_all", is_flag=True, default=False, help="Export all three formats.")
-@click.option("--min-trust", default=0.7, show_default=True, help="Minimum trust score (SFT/reward).")
+@click.option(
+    "--min-trust", default=0.7, show_default=True, help="Minimum trust score (SFT/reward)."
+)
 @click.option("--topic", default=None, help="Filter by topic.")
-@click.option("--exclude-contested", is_flag=True, default=False, help="Exclude claims with open disputes.")
-@click.option("--include-merkle-proof", is_flag=True, default=False, help="Add Merkle proof per record.")
 @click.option(
-    "--output-format", type=click.Choice(["jsonl", "parquet", "hf"]), default="jsonl", show_default=True,
-    help="Output format: jsonl (default), parquet (needs pyarrow), hf (needs datasets)."
+    "--exclude-contested", is_flag=True, default=False, help="Exclude claims with open disputes."
+)
+@click.option(
+    "--include-merkle-proof", is_flag=True, default=False, help="Add Merkle proof per record."
+)
+@click.option(
+    "--output-format",
+    type=click.Choice(["jsonl", "parquet", "hf"]),
+    default="jsonl",
+    show_default=True,
+    help="Output format: jsonl (default), parquet (needs pyarrow), hf (needs datasets).",
+)
+@click.option(
+    "--push-to-hub", default=None, help="HuggingFace dataset repo (e.g. my-org/my-dataset)."
 )
-@click.option("--push-to-hub", default=None, help="HuggingFace dataset repo (e.g. my-org/my-dataset).")
 @click.option("--swarm", default=3, show_default=True)
 def dataset_export_cmd(
     fmt: str,
@@ -1804,7 +1860,8 @@ def dataset_export_cmd(
             dest_name = f"{f}.{ext}" if ext else f"{f}_hf"
             dest = out if (not export_all and out) else os.path.join(base_dir, dest_name)
             result = await export_dataset(
-                hm, f,
+                hm,
+                f,
                 min_trust=min_trust,
                 topic=topic,
                 out=dest,

{hypermind-1.4.6 → hypermind-1.4.7}/src/hypermind/__init__.py

@@ -73,7 +73,7 @@ open = HyperMindAgent.open
 #: is held until the §External & human gates clear (IETF I-D, IANA,
 #: third-party audit, OpenSSF acceptance, ≥3 pilots) — at that point we
 #: cut ``1.0.0`` GA. See ROADMAP.md §Status snapshot.
-__version__ = "1.4.6"
+__version__ = "1.4.7"
 
 #: SLSA build-provenance level committed for every release artefact.
 #: Cosign keyless signing + slsa-github-generator + PEP 740 attestations

{hypermind-1.4.6 → hypermind-1.4.7}/src/hypermind/dataset.py

@@ -155,14 +155,20 @@ async def export_dataset(
         world = getattr(client._primary, "_world", None)
         if world:
             from hypermind.dispute import DisputeStatus
+
             for kid, d in getattr(world, "disputes", {}).items():
                 state = getattr(d, "state", None)
                 if state and state not in (DisputeStatus.CLOSED, DisputeStatus.FRIVOLOUS):
                     kid_hex = kid.hex() if isinstance(kid, (bytes, bytearray)) else str(kid)
                     contested_kids.add(kid_hex)
         claims_raw = [
-            c for c in claims_raw
-            if (bytes(c["kid"]).hex() if isinstance(c.get("kid"), (bytes, bytearray)) else str(c.get("kid", "")))
+            c
+            for c in claims_raw
+            if (
+                bytes(c["kid"]).hex()
+                if isinstance(c.get("kid"), (bytes, bytearray))
+                else str(c.get("kid", ""))
+            )
             not in contested_kids
         ]
 
@@ -243,9 +249,7 @@ async def export_dataset(
                 import pyarrow as pa  # type: ignore[import]
                 import pyarrow.parquet as pq  # type: ignore[import]
             except ImportError as exc:
-                raise ImportError(
-                    "Parquet output requires pyarrow: pip install pyarrow"
-                ) from exc
+                raise ImportError("Parquet output requires pyarrow: pip install pyarrow") from exc
             table = pa.Table.from_pylist(records)
             pq.write_table(table, out_path)
         elif output_format == "hf":

{hypermind-1.4.6 → hypermind-1.4.7}/src/hypermind/integrations/client.py

@@ -137,6 +137,7 @@ class HyperMindClient:
         ``hm_related`` can traverse the citation graph without RELATA_URL.
         """
         import time as _time
+
         topic = topic or self._topic
         unc = confidence_to_uncertainty(confidence, effective_n=10.0)
         kid = await self._primary.publish_claim({"finding": text}, uncertainty=unc, topic=topic)

{hypermind-1.4.6 → hypermind-1.4.7}/src/hypermind/integrations/contradiction.py

@@ -81,13 +81,25 @@ def scan_contradictions(
             def _hex(k: Any) -> str:
                 return k.hex() if isinstance(k, (bytes, bytearray)) else str(k)
 
-            pairs.append({
-                "claim_a": {"kid": _hex(a_kid), "text": a.get("text", "")[:200], "confidence": a.get("confidence"), "topic": a.get("topic")},
-                "claim_b": {"kid": _hex(b_kid), "text": b.get("text", "")[:200], "confidence": b.get("confidence"), "topic": b.get("topic")},
-                "similarity": round(sim, 4),
-                "conflict_type": conflict_type,
-                "conflict_score": conflict_score,
-            })
+            pairs.append(
+                {
+                    "claim_a": {
+                        "kid": _hex(a_kid),
+                        "text": a.get("text", "")[:200],
+                        "confidence": a.get("confidence"),
+                        "topic": a.get("topic"),
+                    },
+                    "claim_b": {
+                        "kid": _hex(b_kid),
+                        "text": b.get("text", "")[:200],
+                        "confidence": b.get("confidence"),
+                        "topic": b.get("topic"),
+                    },
+                    "similarity": round(sim, 4),
+                    "conflict_type": conflict_type,
+                    "conflict_score": conflict_score,
+                }
+            )
             if len(pairs) >= max_pairs:
                 break