The latest articles on DEV Community by BerSerCarrrrrrrrrrr (@ruanflux). https://dev.to/ruanflux https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3888389%2Fdc01bee9-9ed5-48db-b097-f688ad64ee9e.png https://dev.to/ruanflux en BerSerCarrrrrrrrrrr Mon, 20 Apr 2026 07:03:13 +0000 https://dev.to/ruanflux/i-built-chatrel-applying-7-psychology-frameworks-to-your-chat-history-46j6 https://dev.to/ruanflux/i-built-chatrel-applying-7-psychology-frameworks-to-your-chat-history-46j6 <p>Most chat analyzers give you a word cloud, a sentiment line, and a heatmap. <br> I wanted something different: a local tool that takes a chat history with <br> a partner and maps it onto frameworks that relationship scientists have <br> actually studied for 40+ years.</p> <p>I spent a weekend building <strong><a href="https://github.com/ruanflux/chatrel" rel="noopener noreferrer">ChatRel</a></strong>.</p> <p>🚀 <strong>Live demo (no install, synthetic data)</strong>: <a href="https://ruanflux.github.io/chatrel/" rel="noopener noreferrer">https://ruanflux.github.io/chatrel/</a></p> <h2> The 7 frameworks (all with citations) </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Framework</th> <th>Measures</th> <th>Source</th> </tr> </thead> <tbody> <tr> <td><strong>Sternberg Triangular</strong></td> <td>Intimacy × Passion × Commitment, weekly</td> <td>Sternberg (1986)</td> </tr> <tr> <td><strong>Knapp Stages</strong></td> <td>10-stage relational development</td> <td>Knapp (1978)</td> </tr> <tr> <td><strong>ECR-R Attachment</strong></td> <td>Anxiety × Avoidance, Bartholomew 4-category</td> <td>Fraley et al. (2000)</td> </tr> <tr> <td><strong>Gottman 5:1 + Four Horsemen</strong></td> <td>Positive-negative ratio, criticism/contempt/defensiveness/stonewalling</td> <td>Gottman (1994)</td> </tr> <tr> <td><strong>Social Penetration</strong></td> <td>Self-disclosure depth</td> <td>Altman &amp; Taylor (1973)</td> </tr> <tr> <td><strong>Critical Slowing (EWS)</strong></td> <td>Variance + AR(1) as regime-shift warnings</td> <td>Scheffer et al. (2009)</td> </tr> <tr> <td><strong>LIWC (Chinese proxy)</strong></td> <td>Function-word / cognitive / affect markers</td> <td>Pennebaker</td> </tr> </tbody> </table></div> <h2> What you get </h2> <p>Seven interactive HTML reports:</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fd1kzxp8l67jhklnabmbx.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fd1kzxp8l67jhklnabmbx.png" alt="Depth report" width="800" height="562"></a></p> <ul> <li>Weekly <strong>IDS</strong> (Intimacy Depth Score) with Sternberg components decomposed</li> <li> <strong>Attachment 2D scatter</strong> placing you and your partner on anxiety × avoidance</li> <li> <strong>Conflict-repair events</strong> with initiator, peak, apology lag, and resolution</li> <li> <strong>External resilience</strong> score (how well the relationship handles poaching / ex-mentions)</li> <li> <strong>Memory book</strong> — auto-curated firsts, sweetest, and deepest moments</li> <li> <strong>12-week forecast</strong> (base / optimistic / pessimistic scenarios)</li> </ul> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F5udsn00tq38jbd1fvm3t.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F5udsn00tq38jbd1fvm3t.png" alt="Attachment scatter" width="800" height="562"></a></p> <h2> Privacy is the architecture, not a feature </h2> <ul> <li> <strong>100% local</strong>. No API calls. No telemetry.</li> <li>Supports Telegram JSON / generic CSV / WeChat (via external decryption)</li> <li>Ships with a synthetic 90-day demo chat — you can evaluate the whole tool without real data</li> <li> <code>.gitignore</code> blocks every pattern that would leak chat content</li> </ul> <h2> Known limits (documented honestly) </h2> <ul> <li> <strong>n=1 validation</strong>. Nothing calibrated against ECR-R self-report scores. Proxy signals, not diagnoses.</li> <li>Forecasts are <strong>damped linear extrapolation with mean reversion</strong>, not Bayesian.</li> <li>Contempt detection had to exclude <code>笑死</code> (modern Chinese "lmao", not contempt) — similar traps are in-file documented.</li> <li>Lexicons are Chinese-only. PRs for other languages welcome.</li> </ul> <h2> Stack </h2> <p>Python 3.10+ · <code>jieba</code> · <code>snownlp</code> · <code>pandas</code> · <code>plotly</code>. MIT licensed.</p> <h2> Try it </h2> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> bash git clone https://github.com/ruanflux/chatrel.git cd chatrel pip install -e . chatrel sample --out sample_data/synthetic_couple.json chatrel analyze sample_data/synthetic_couple.json --out ./output </code></pre> </div> opensource python psychology nlp