dialectical-framework 1.6.2__tar.gz → 1.6.4__tar.gz

dialectical_framework-1.6.4/PATENTS

@@ -0,0 +1,79 @@
+Dialectical Framework
+Copyright (c) 2025 Dialexity
+
+================================================================================
+COPYRIGHT LICENSE
+================================================================================
+
+The source code in this repository is licensed under the MIT License. See the
+LICENSE file for the full terms. The MIT License governs COPYRIGHT ONLY: your
+rights to use, copy, modify, merge, publish, distribute, sublicense, and sell
+copies of the software.
+
+================================================================================
+PATENT NOTICE
+================================================================================
+
+Independently of the copyright license above, Dialexity holds patent rights
+covering methods for applying dialectical reasoning with artificial
+intelligence and machine learning systems.
+
+  Patent(s) / application(s): see https://dialexity.com/patents
+
+The MIT License does not, by itself, grant any patent rights. This file sets
+out the patent terms that apply to this software.
+
+--------------------------------------------------------------------------------
+1. Limited patent grant for use of this software as distributed
+--------------------------------------------------------------------------------
+
+Dialexity grants you a worldwide, royalty-free, non-exclusive license under the
+patent(s) identified above, limited strictly to making, using, selling,
+offering for sale, importing, and distributing THIS software as distributed in
+this repository (and modifications of it), to the extent such acts would
+otherwise infringe those patent(s).
+
+This grant is coextensive with your use of this software. It extends ONLY to
+the patented methods as embodied in this software and does NOT extend to any
+other implementation of the patented methods.
+
+--------------------------------------------------------------------------------
+2. Reservation of all other patent rights
+--------------------------------------------------------------------------------
+
+All patent rights not expressly granted in Section 1 are reserved. In
+particular, and without limiting the foregoing, NO patent license is granted
+for:
+
+  (a) implementations of the patented methods that are developed independently
+      of, or do not incorporate, this software; or
+
+  (b) products, services, or systems that practice the patented methods without
+      using this software as distributed.
+
+Practicing the patented methods in any manner outside the limited grant in
+Section 1 requires a separate patent license from Dialexity.
+
+--------------------------------------------------------------------------------
+3. Defensive termination
+--------------------------------------------------------------------------------
+
+The patent grant in Section 1 terminates automatically as to any party that
+initiates or joins a patent infringement claim (including a cross-claim or
+counterclaim) alleging that this software, or any contribution to it,
+constitutes patent infringement.
+
+--------------------------------------------------------------------------------
+4. Commercial and patent licensing
+--------------------------------------------------------------------------------
+
+For patent licenses beyond the limited grant in Section 1 — including use of
+the patented methods in independent implementations — see the patent listing
+and contact details at:
+
+  https://dialexity.com/patents
+
+================================================================================
+This file must be included in all copies or substantial portions of the
+software, together with the LICENSE file.
+================================================================================

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: dialectical-framework
-Version: 1.6.2
+Version: 1.6.4
 Summary: A dialectical framework for augmented intelligence. AI reasoning powered with dialectics supports humans in: system optimization (psychology, engineering, business, politics, etc.); dispute resolution (mediation, conflicts, negotiations, etc.); decision-making (dilemmas, challenging situations, win-win, etc.).
 License: MIT
 Keywords: dialectics,dialectical-reasoning,synthesis,thesis-antithesis,ai,artificial-intelligence,llm,reasoning-framework,philosophy,logic,argumentation,conflict-resolution,decision-making,critical-thinking,semantic-graph,mirascope,pydantic,perspectives,polarity-reasoning
@@ -176,3 +176,12 @@ poetry run pytest --real-llm   # Hit real LLM provider
 - [Mirascope](https://mirascope.com/) — LLM abstraction
 - [GQLAlchemy](https://memgraph.com/docs/gqlalchemy) — Graph ORM
 - [dependency-injector](https://python-dependency-injector.ets-labs.org/) — DI container
+
+## License & Patents
+
+The source code is licensed under the [MIT License](https://github.com/dialexity/dialectical-framework/blob/main/LICENSE) (copyright only).
+
+The methods for applying dialectical reasoning with AI are patented. The MIT
+License does not grant patent rights beyond a limited grant for use of this
+software as distributed — see [PATENTS](https://github.com/dialexity/dialectical-framework/blob/main/PATENTS) for the patent terms and
+[dialexity.com/patents](https://dialexity.com/patents) for the patent listing.

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/README.md

@@ -134,4 +134,13 @@ poetry run pytest --real-llm   # Hit real LLM provider
 
 - [Mirascope](https://mirascope.com/) — LLM abstraction
 - [GQLAlchemy](https://memgraph.com/docs/gqlalchemy) — Graph ORM
-- [dependency-injector](https://python-dependency-injector.ets-labs.org/) — DI container
+- [dependency-injector](https://python-dependency-injector.ets-labs.org/) — DI container
+
+## License & Patents
+
+The source code is licensed under the [MIT License](https://github.com/dialexity/dialectical-framework/blob/main/LICENSE) (copyright only).
+
+The methods for applying dialectical reasoning with AI are patented. The MIT
+License does not grant patent rights beyond a limited grant for use of this
+software as distributed — see [PATENTS](https://github.com/dialexity/dialectical-framework/blob/main/PATENTS) for the patent terms and
+[dialexity.com/patents](https://dialexity.com/patents) for the patent listing.

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "dialectical-framework"
-version = "1.6.2"
+version = "1.6.4"
 description = "A dialectical framework for augmented intelligence. AI reasoning powered with dialectics supports humans in: system optimization (psychology, engineering, business, politics, etc.); dispute resolution (mediation, conflicts, negotiations, etc.); decision-making (dilemmas, challenging situations, win-win, etc.)."
 authors = ["Evaldas Taroza <evaldas@dialexity.com>"]
 readme = "README.md"
@@ -35,6 +35,7 @@ classifiers = [
     "Environment :: Console",
     "Natural Language :: English"
 ]
+include = [{ path = "PATENTS", format = ["sdist"] }]
 
 [[tool.poetry.packages]]
 include = "dialectical_framework"

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/agents/analyst/analyst.py

@@ -318,6 +318,14 @@ class AnalysisPipeline(ReasonableConcern[AnalysisResult]):
                 seen.add(h)
                 hashes_to_expand.append(h)
 
+        # Surface the HS-based gate to the agent. _rank_polarities silently
+        # drops tensions below HS_THRESHOLD (and beyond the top few); without
+        # this the agent only sees a bare perspective count and cannot explain
+        # WHY these framings are the strong ones or that weaker tensions were
+        # set aside.
+        polarity_quality = self._build_polarity_quality(polarity_data, hashes_to_expand)
+        set_aside_count = sum(1 for q in polarity_quality if not q["expanded"])
+
         expand_results = await asyncio.gather(
             *[self._expand_one(h) for h in hashes_to_expand],
             return_exceptions=True,
@@ -338,14 +346,21 @@ class AnalysisPipeline(ReasonableConcern[AnalysisResult]):
                 reports.append(report)
 
         self._report.ok = True
+        set_aside_note = (
+            f", {set_aside_count} weaker tension(s) set aside"
+            if set_aside_count
+            else ""
+        )
         self._report.summary = (
             f"Analysis complete: {len(thesis_hashes)} theses, "
             f"{len(polarity_hashes)} polarities, "
             f"{len(perspective_hashes)} perspectives"
+            f"{set_aside_note}"
         )
         self._report.artifacts["thesis_hashes"] = thesis_hashes
         self._report.artifacts["polarity_hashes"] = polarity_hashes
         self._report.artifacts["perspective_hashes"] = perspective_hashes
+        self._report.artifacts["polarity_quality"] = polarity_quality
 
         return AnalysisResult(
             ideas_hash=ideas_hash,
@@ -375,6 +390,39 @@ class AnalysisPipeline(ReasonableConcern[AnalysisResult]):
 
         return ranked[:MAX_POLARITIES_TO_EXPAND]
 
+    @staticmethod
+    def _build_polarity_quality(
+        polarity_data: list[dict], hashes_to_expand: list[str]
+    ) -> list[dict]:
+        """Per-tension HS + gate outcome, HS-descending, for the agent to read.
+
+        Makes the otherwise-silent HS gate visible: each entry flags whether the
+        tension was `expanded` into a full perspective or set aside. HS on the
+        antithesis measures how genuine the opposition is (see "Reading Polarity
+        Quality" in the system prompt), not the quality of the idea itself.
+        """
+        expanded_set = set(hashes_to_expand)
+        quality: list[dict] = []
+        seen: set[str] = set()
+        for p in polarity_data:
+            h = p.get("polarity_hash")
+            if not h or h in seen or p.get("deduped", False):
+                continue
+            seen.add(h)
+            quality.append(
+                {
+                    "polarity_hash": h,
+                    "thesis": p.get("thesis_text"),
+                    "antithesis": p.get("antithesis_text"),
+                    "hs": p.get("heuristic_similarity"),
+                    "expanded": h in expanded_set,
+                }
+            )
+        quality.sort(
+            key=lambda q: q["hs"] if q["hs"] is not None else 0.0, reverse=True
+        )
+        return quality
+
     async def _expand_one(self, polarity_hash: str) -> tuple[list[str], object]:
         from dialectical_framework.agents.analyst.skills.expand_polarities import \
             ExpandPolarity

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/agents/analyst/skills/find_polarities.py

@@ -39,8 +39,13 @@ from dialectical_framework.agents.execution_report import ExecutionReport
 from dialectical_framework.enums.di import DI
 from dialectical_framework.concerns.antithesis_extraction import \
     AntithesisExtraction
+from dialectical_framework.concerns.antithetical_thesis_detection import \
+    AntitheticalThesisDetection
 from dialectical_framework.concerns.statement_deduplication import (
     DedupResult, StatementDeduplication)
+from dialectical_framework.graph.estimation_manager import EstimationManager
+from dialectical_framework.graph.nodes.estimation import (ArousalEstimation,
+                                                          ModeEstimation)
 from dialectical_framework.graph.nodes.statement import \
     Statement
 from dialectical_framework.graph.nodes.ideas import Ideas
@@ -121,6 +126,15 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
         results: list[ThesisResult] = []
         newly_extracted: list[Statement] = []
 
+        # Phase 0: Consolidate theses that are antitheses of each other. Strongly
+        # opposed pairs (HS >= 0.7) are merged into a single Polarity here and
+        # removed from extraction; weaker pairs are surfaced as suggestions for
+        # the agent to confirm (leaving both theses to proceed independently).
+        unique_hashes = list(dict.fromkeys(self.thesis_hashes))
+        unique_hashes, consolidated_data = await self._consolidate_antithetical(
+            unique_hashes, input_text
+        )
+
         # Phase 1: For each thesis, collect existing oppositions + extract new ones (parallel)
         import asyncio
 
@@ -155,7 +169,6 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
             result.existing = existing_antitheses
             return result
 
-        unique_hashes = list(dict.fromkeys(self.thesis_hashes))
         thesis_results = await asyncio.gather(
             *[_process_thesis(h) for h in unique_hashes]
         )
@@ -209,18 +222,20 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
         # Phase 3: Create Polarity nodes for each T-A pair
         total_antitheses = sum(len(r.antithesis_data) for r in results if not r.error)
 
-        if total_antitheses == 0:
+        if total_antitheses == 0 and not consolidated_data:
             self._report.ok = True
             self._report.summary = "No polarities found"
             self._report.artifacts["polarity_data"] = []
             return None
 
-        # Phase 4: Create Polarities and Ideas with all T-A pairs
-        ideas = self._create_ideas(results)
+        # Phase 4: Create Polarities and Ideas with all T-A pairs (consolidated
+        # pairs already have their Polarities created in Phase 0).
+        ideas = self._create_ideas(results, consolidated_data)
         polarity_map = self._create_polarities(results)
 
-        # Build polarity_data for report (includes HS for each T-A pair)
-        polarity_data = []
+        # Build polarity_data for report (includes HS for each T-A pair). Start
+        # with the consolidated pairs, then the extracted ones.
+        polarity_data = list(consolidated_data)
         for result in results:
             if result.error:
                 continue
@@ -258,11 +273,26 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
         self._report.artifacts["polarity_data"] = polarity_data
         pol_created = self._report.artifacts.get("created_polarity_count", 0)
         pol_existing = self._report.artifacts.get("existing_polarity_count", 0)
-        self._report.summary = (
+        consolidated_count = len(consolidated_data)
+        suggestion_count = len(
+            self._report.artifacts.get("consolidation_suggestions", [])
+        )
+        summary = (
             f"Found {existing_count} existing + {new_count} new antithesis(es) "
             f"for {len(self.thesis_hashes)} thesis(es). "
             f"Polarities: {pol_created} created, {pol_existing} existing."
         )
+        if consolidated_count:
+            summary += (
+                f" Consolidated {consolidated_count} antithetical thesis pair(s) "
+                f"into single tension(s)."
+            )
+        if suggestion_count:
+            summary += (
+                f" {suggestion_count} weaker pair(s) suggested for consolidation "
+                f"(acting on one merges 2 theses into 1 tension)."
+            )
+        self._report.summary = summary
 
         return ideas
 
@@ -424,6 +454,115 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
             pass
         return None
 
+    async def _consolidate_antithetical(
+        self, unique_hashes: list[str], input_text: str
+    ) -> tuple[list[str], list[dict]]:
+        """Detect and merge theses that are antitheses of each other.
+
+        Strongly-opposed pairs (HS >= 0.7) are merged into a single Polarity and
+        their two hashes removed from the returned list (so they are NOT sent to
+        antithesis extraction). Weaker pairs (0.1 < HS < 0.7) are recorded as
+        suggestions but left in the list to proceed as independent theses.
+
+        Returns:
+            (reduced_hashes, consolidated_polarity_data) where each dict has the
+            same shape FindPolarities uses in ``polarity_data``.
+        """
+        if len(unique_hashes) < 2:
+            return unique_hashes, []
+
+        detector = AntitheticalThesisDetection()
+        detection = await detector.resolve(
+            thesis_hashes=unique_hashes, text=input_text
+        )
+        self._report = self._report.merge(detector.report)
+
+        # Surface weak pairs for the agent to confirm (non-destructive).
+        self._report.artifacts["consolidation_suggestions"] = [
+            p.as_dict() for p in detection.suggest_pairs
+        ]
+
+        if not detection.merge_pairs:
+            return unique_hashes, []
+
+        pol_repo = PolarityRepository()
+        est_manager = EstimationManager()
+        consolidated_data: list[dict] = []
+        merged_hashes: set[str] = set()
+        merged_count = 0
+
+        for pair in detection.merge_pairs:
+            thesis = self._resolve_component(pair.thesis_hash)
+            antithesis = self._resolve_component(pair.antithesis_hash)
+            if thesis is None or antithesis is None:
+                continue
+
+            # Record the dialectical opposition (symmetric, idempotent).
+            thesis.oppositions.connect(antithesis)
+            self._report.relationship_created(
+                thesis.oppositions, thesis, antithesis
+            )
+
+            existing_pols = pol_repo.find_by_tension(thesis, antithesis)
+            if existing_pols:
+                polarity = existing_pols[0]
+                self._report.artifacts.setdefault("existing_polarity_count", 0)
+                self._report.artifacts["existing_polarity_count"] += 1
+            else:
+                polarity = self._build_polarity(
+                    thesis, antithesis, pair.heuristic_similarity
+                )
+                merged_count += 1
+
+            # Persist antithesis Mode/Arousal (Antithesis Persistence Checklist).
+            self._persist_mode_arousal(
+                est_manager, antithesis, pair.mode_value, pair.arousal_value
+            )
+
+            consolidated_data.append(
+                {
+                    "polarity_hash": polarity.hash,
+                    "thesis_hash": thesis.hash,
+                    "thesis_text": thesis.text,
+                    "antithesis_hash": antithesis.hash,
+                    "antithesis_text": antithesis.text,
+                    "heuristic_similarity": pair.heuristic_similarity,
+                    "existing": False,
+                    "deduped": False,
+                    "consolidated": True,
+                }
+            )
+            merged_hashes.add(thesis.hash)
+            merged_hashes.add(antithesis.hash)
+
+        self._report.artifacts["consolidated_pairs"] = len(consolidated_data)
+        if merged_count:
+            self._report.artifacts["created_polarity_count"] = (
+                self._report.artifacts.get("created_polarity_count", 0) + merged_count
+            )
+
+        reduced = [h for h in unique_hashes if h not in merged_hashes]
+        return reduced, consolidated_data
+
+    def _persist_mode_arousal(
+        self,
+        est_manager: EstimationManager,
+        antithesis: Statement,
+        mode_value: float,
+        arousal_value: float,
+    ) -> None:
+        """Persist Mode/Arousal estimations on the antithesis (mirrors IntroducePolarity)."""
+        mode_est = est_manager.upsert_estimation(
+            antithesis, ModeEstimation, mode_value
+        )
+        arousal_est = est_manager.upsert_estimation(
+            antithesis, ArousalEstimation, arousal_value
+        )
+        if mode_est:
+            self._report.node_updated(mode_est, patch={"value": mode_value})
+        if arousal_est:
+            self._report.node_updated(arousal_est, patch={"value": arousal_value})
+
     def _create_polarities(self, results: list[ThesisResult]) -> dict[tuple[str, str], str]:
         """Create Polarity nodes (T-A pairs) for each T-A pair.
 
@@ -461,37 +600,63 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
                     continue
 
                 # Create new Polarity (atomic creation)
-                polarity = Polarity()
-                polarity.set_t(result.thesis, heuristic_similarity=1.0)
-                polarity.set_a(
-                    antithesis, heuristic_similarity=data["heuristic_similarity"]
+                polarity = self._build_polarity(
+                    result.thesis, antithesis, data["heuristic_similarity"]
                 )
-                polarity.commit()
-
                 polarity_map[(result.thesis.hash, data["hash"])] = polarity.hash
                 created_count += 1
-                self._report.node_created(
-                    polarity, meta={"hs": data["heuristic_similarity"]}
-                )
-                self._report.relationship_created(
-                    polarity.t, result.thesis, polarity,
-                    patch={"heuristic_similarity": 1.0, "alias": "T"},
-                )
-                self._report.relationship_created(
-                    polarity.a, antithesis, polarity,
-                    patch={"heuristic_similarity": data["heuristic_similarity"], "alias": "A"},
-                )
 
-        self._report.artifacts["created_polarity_count"] = created_count
+        # Accumulate — Phase 0 consolidation may already have created polarities.
+        self._report.artifacts["created_polarity_count"] = (
+            self._report.artifacts.get("created_polarity_count", 0) + created_count
+        )
         return polarity_map
 
-    def _create_ideas(self, results: list[ThesisResult]) -> Optional[Ideas]:
-        """Create Ideas node with all theses and their antitheses."""
+    def _build_polarity(
+        self,
+        thesis: Statement,
+        antithesis: Statement,
+        heuristic_similarity: float,
+    ) -> Polarity:
+        """Create and commit a Polarity from two committed Statements, recording effects.
+
+        Shared by regular polarity creation and antithetical-thesis consolidation.
+        T defines the apex (HS=1.0); the meaningful HS lives on A.
+        """
+        polarity = Polarity()
+        polarity.set_t(thesis, heuristic_similarity=1.0)
+        polarity.set_a(antithesis, heuristic_similarity=heuristic_similarity)
+        polarity.commit()
+
+        self._report.node_created(polarity, meta={"hs": heuristic_similarity})
+        self._report.relationship_created(
+            polarity.t, thesis, polarity,
+            patch={"heuristic_similarity": 1.0, "alias": "T"},
+        )
+        self._report.relationship_created(
+            polarity.a, antithesis, polarity,
+            patch={"heuristic_similarity": heuristic_similarity, "alias": "A"},
+        )
+        return polarity
+
+    def _create_ideas(
+        self,
+        results: list[ThesisResult],
+        consolidated_data: Optional[list[dict]] = None,
+    ) -> Optional[Ideas]:
+        """Create Ideas node with all theses and their antitheses.
+
+        Includes both the extracted T-A pairs (from ``results``) and the
+        consolidated antithetical pairs (from ``consolidated_data``), whose
+        theses were removed from the extraction set in Phase 0.
+        """
+        consolidated_data = consolidated_data or []
         valid_results = [r for r in results if not r.error and r.antithesis_data]
-        if not valid_results:
+        if not valid_results and not consolidated_data:
             return None
 
         thesis_statements = [r.thesis.text for r in valid_results]
+        thesis_statements += [d["thesis_text"] for d in consolidated_data]
         intent = f"Tensions for: {', '.join(thesis_statements[:3])}"
         if len(thesis_statements) > 3:
             intent += f" (+{len(thesis_statements) - 3} more)"
@@ -508,33 +673,38 @@ class FindPolarities(ReasonableConcern[Optional[Ideas]]):
 
         # Connect all theses and antitheses (deduplicated across theses)
         connected_hashes: set[str] = set()
+
+        def _connect(hash: str) -> None:
+            if hash in connected_hashes:
+                return
+            comp = self._resolve_component(hash)
+            if comp:
+                connected_hashes.add(hash)
+                ideas.statements.connect(comp)
+                self._report.relationship_created(ideas.statements, ideas, comp)
+
         for result in valid_results:
-            if result.thesis.hash not in connected_hashes:
-                connected_hashes.add(result.thesis.hash)
-                ideas.statements.connect(result.thesis)
-                self._report.relationship_created(
-                    ideas.statements, ideas, result.thesis
-                )
+            _connect(result.thesis.hash)
             for data in result.antithesis_data:
-                if data["hash"] in connected_hashes:
-                    continue
-                comp = self._resolve_component(data["hash"])
-                if comp:
-                    connected_hashes.add(data["hash"])
-                    ideas.statements.connect(comp)
-                    self._report.relationship_created(
-                        ideas.statements, ideas, comp
-                    )
+                _connect(data["hash"])
+
+        # Consolidated pairs: both sides are theses the user supplied.
+        for data in consolidated_data:
+            _connect(data["thesis_hash"])
+            _connect(data["antithesis_hash"])
 
         ideas.commit()
         self._report.node_committed(ideas)
 
         # Add rationale
-        total_theses = len(valid_results)
-        total_antitheses = sum(len(r.antithesis_data) for r in valid_results)
+        total_theses = len(valid_results) + len(consolidated_data)
+        total_antitheses = sum(len(r.antithesis_data) for r in valid_results) + len(
+            consolidated_data
+        )
         all_hs = [
             d["heuristic_similarity"] for r in valid_results for d in r.antithesis_data
         ]
+        all_hs += [d["heuristic_similarity"] for d in consolidated_data]
         max_hs = max(all_hs) if all_hs else 0.0
 
         rationale = Rationale(

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/agents/analyst/system_prompts.py

@@ -67,6 +67,43 @@ Always check resonance AFTER presenting results — but never before acting.
 When new tensions emerge from conversation:
 - Call `analyze` with `thesis_hashes` to develop them without re-processing everything.
 
+When `find_polarities` merges antithetical theses (a `polarity_data` entry with
+`consolidated: true`): two theses the user supplied were strong direct opposites,
+so they were combined into one tension automatically. Briefly state that you did
+this — e.g. "I treated X and Y as one tension since they're direct opposites; say
+so if you'd rather explore them separately." Don't ask first, but make the merge
+visible so the user can reverse it (`discard` the polarity and re-anchor them
+apart) if the framing was wrong.
+
+When `find_polarities` returns `consolidation_suggestions` (theses that may be
+antitheses of each other but the opposition is only moderate): present the
+suggested pair(s) to the user and ask whether the two are really one tension. If
+they confirm, call `introduce_polarity` with that thesis/antithesis pair and
+`discard` the now-redundant separate polarities. If they want the theses kept
+apart, leave them as-is. (Strong opposites are merged automatically, so
+suggestions are only the ambiguous middle ground — always confirm before merging.)
+
+## Reading Polarity Quality (HS)
+
+`analyze` and `find_polarities` score every tension with an HS (heuristic
+similarity) on its antithesis — how *genuine* the opposition is, not how good the
+idea is. Use it to decide what to lead with; this is the reasoning that separates
+you from a generic list of pros and cons.
+
+- **≥0.7** — a strong, real opposition. Lead with these.
+- **0.5-0.7** — a moderate or partial tension. Worth presenting, but frame it as secondary.
+- **<0.5** — weak or tangential; the two sides barely oppose. Don't present it as a core tension.
+
+`analyze` expands only the strongest tensions into full perspectives and sets the
+rest aside — the report's `polarity_quality` artifact lists every tension with its
+`hs` and an `expanded` flag, and the summary notes how many were set aside. When
+tensions were set aside, briefly say stronger framings were prioritized and that
+you can develop a weaker one on request (via `find_polarities`/`expand_polarities`
+on that specific tension) — don't silently drop them.
+
+Communicate quality as meaning ("a strong tension", "a weaker, more tangential
+pull"), not raw numbers, unless the app preamble or the user asks for the figures.
+
 ## Exploration Setup (Nexus)
 
 When the user wants to explore interactions between perspectives:

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/agents/apps.py

@@ -15,7 +15,7 @@ They are injected by the host application at agent construction time.
 
 Prompt Revision Methodology
 ===========================
-When fixing LLM output bugs in these prompts, use /df:review-prompts for the
+When fixing LLM output bugs in these prompts, use /df-review-reasoning-layer for the
 full diagnosis/fix/verify workflow.
 
 Quick reference: Diagnose root cause (polysemy | competing signals | missing

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/concerns/ac_re_taxonomy.py

@@ -3,8 +3,6 @@ Ac-Re Taxonomy constants for Transformation generation.
 
 This module contains the scales, polar pairs, and apex target coordinates
 for the Action-Reflection dialectical structure.
-
-See docs/r&d/ac_re-taxonomy.md for full specification.
 """
 
 from __future__ import annotations

{dialectical_framework-1.6.2 → dialectical_framework-1.6.4}/src/dialectical_framework/concerns/antithesis_classification.py

@@ -169,7 +169,7 @@ def arousal_label_to_value(label: str) -> float:
 class ContextualizedTaxonomyDto(BaseModel):
     """Contextualized universal taxonomy for a thesis.
 
-     Structure (from docs/r&d/taxonomy-universal.md):
+     Structure:
                              APEX: [T]-lessness
                                     │
                  ┌──────────────────┴──────────────────┐