Add Codex Responses support and session debugging

Teach provider config and auth resolution about the Codex Responses
dialect, including user-facing `style = "openai_responses"` with
`dialect = "codex"`. Serialize and parse Responses traffic with
provider-specific reasoning replay, assistant phase metadata, and robust
function-call assembly keyed by `output_index` so streamed tool inputs
survive proxy quirks and empty terminal payloads.

Also persist thinking origins and message metadata across sessions, add
the Anthropic interleaved-thinking header switch, write per-session
debug logs, and improve the TUI and scripts for inspecting tool output
and session costs.

1 files changed, 288 insertions, 0 deletions

diff --git a/tmp/pi_session_cost_analysis.json b/tmp/pi_session_cost_analysis.json
new file mode 100644
index 0000000..cc40030
--- /dev/null
+++ b/tmp/pi_session_cost_analysis.json
@@ -0,0 +1,288 @@
+{
+  "root": "/Users/travis/Code/config/dot-pi/agent/sessions",
+  "sessions_scanned": 495,
+  "sessions_with_usage": 454,
+  "sessions_without_usage": 41,
+  "assistant_messages": 28636,
+  "assistant_messages_with_usage": 28636,
+  "parse_errors": 0,
+  "cache_active_sessions": 216,
+  "pooled_totals": {
+    "input": 648085868,
+    "output": 11386363,
+    "cache_read": 984042157,
+    "cache_write": 65075433
+  },
+  "pooled_metrics": {
+    "prompt_tokens": 1697203458,
+    "billable_tokens": 1708589821,
+    "prompt_to_output_ratio": 149.0558010490268,
+    "input_to_output_ratio": 56.91772412314626,
+    "cache_hit_rate": 0.9379712687878963,
+    "cache_coverage": 0.6181448576803453,
+    "prompt_reuse_rate": 0.5798021164531472,
+    "output_share": 0.006664187542294857,
+    "prompt_share": 0.9933358124577052,
+    "input_share": 0.37931038803724676,
+    "cache_read_share": 0.5759382064116839,
+    "cache_write_share": 0.0380872180087745
+  },
+  "recommended_weights": {
+    "pooled_token_mix": {
+      "input": 0.37931038803724676,
+      "output": 0.006664187542294857,
+      "cache_read": 0.5759382064116839,
+      "cache_write": 0.0380872180087745
+    },
+    "mean_session_mix": {
+      "input": 0.549819370457156,
+      "output": 0.05134274693943982,
+      "cache_read": 0.3229895728469125,
+      "cache_write": 0.07584830975649165
+    },
+    "median_session_mix": {
+      "input": 0.7903769253077484,
+      "output": 0.014961511819985934,
+      "cache_read": 0.0,
+      "cache_write": 0.0
+    }
+  },
+  "per_session_distributions": {
+    "billable_tokens": {
+      "count": 454,
+      "mean": 3763413.702643172,
+      "median": 315231.0,
+      "p10": 20182.100000000002,
+      "p90": 11804297.1,
+      "min": 2244,
+      "max": 101170942
+    },
+    "cache_coverage": {
+      "count": 454,
+      "mean": 0.40473111896757075,
+      "median": 0.0,
+      "p10": 0.0,
+      "p90": 0.9999679722217543,
+      "min": 0.0,
+      "max": 0.9999922748401895
+    },
+    "cache_hit_rate": {
+      "count": 216,
+      "mean": 0.8218859540276632,
+      "median": 0.9487389145525554,
+      "p10": 0.374890408374274,
+      "p90": 1.0,
+      "min": 0.0,
+      "max": 1.0
+    },
+    "cache_read_share": {
+      "count": 454,
+      "mean": 0.3229895728469125,
+      "median": 0.0,
+      "p10": 0.0,
+      "p90": 0.9389935967096287,
+      "min": 0.0,
+      "max": 0.9893885889087582
+    },
+    "cache_write_share": {
+      "count": 454,
+      "mean": 0.07584830975649165,
+      "median": 0.0,
+      "p10": 0.0,
+      "p90": 0.267090715504709,
+      "min": 0.0,
+      "max": 0.999839897534422
+    },
+    "input_share": {
+      "count": 454,
+      "mean": 0.549819370457156,
+      "median": 0.7903769253077484,
+      "p10": 3.1707600618505675e-05,
+      "p90": 0.9974440745899954,
+      "min": 7.709723608187813e-06,
+      "max": 0.9999909002957817
+    },
+    "input_to_output_ratio": {
+      "count": 454,
+      "mean": 829.6580689032315,
+      "median": 8.629654523169638,
+      "p10": 0.00421753985473397,
+      "p90": 409.9444444444443,
+      "min": 0.0010703582132153562,
+      "max": 109892.68181818182
+    },
+    "output_share": {
+      "count": 454,
+      "mean": 0.05134274693943982,
+      "median": 0.014961511819985934,
+      "p10": 0.0018531251145351706,
+      "p90": 0.16915366650275257,
+      "min": 9.099704218250615e-06,
+      "max": 0.5140911569511498
+    },
+    "output_tokens": {
+      "count": 454,
+      "mean": 25080.09471365639,
+      "median": 7470.5,
+      "p10": 365.8000000000001,
+      "p90": 78289.59999999998,
+      "min": 1,
+      "max": 279572
+    },
+    "prompt_reuse_rate": {
+      "count": 454,
+      "mean": 0.32683124761341736,
+      "median": 0.0,
+      "p10": 0.0,
+      "p90": 0.9461329490069706,
+      "min": 0.0,
+      "max": 0.9914821057091003
+    },
+    "prompt_share": {
+      "count": 454,
+      "mean": 0.9486572530605601,
+      "median": 0.985038488180014,
+      "p10": 0.8308463334972475,
+      "p90": 0.9981468748854648,
+      "min": 0.48590884304885024,
+      "max": 0.9999909002957817
+    },
+    "prompt_to_output_ratio": {
+      "count": 454,
+      "mean": 938.8894985366835,
+      "median": 65.83823653274615,
+      "p10": 4.91178438425,
+      "p90": 538.8489050453238,
+      "min": 0.9451803176902779,
+      "max": 109892.68181818182
+    },
+    "prompt_tokens": {
+      "count": 454,
+      "mean": 3738333.607929515,
+      "median": 308096.0,
+      "p10": 19041.5,
+      "p90": 11756748.7,
+      "min": 2149,
+      "max": 100968785
+    }
+  },
+  "cache_active_session_distributions": {
+    "cache_coverage": {
+      "count": 216,
+      "mean": 0.8506848519040607,
+      "median": 0.9997448273674421,
+      "p10": 0.43102818092852674,
+      "p90": 0.999982772379876,
+      "min": 0.0021478132287605567,
+      "max": 0.9999922748401895
+    },
+    "cache_hit_rate": {
+      "count": 216,
+      "mean": 0.8218859540276632,
+      "median": 0.9487389145525554,
+      "p10": 0.374890408374274,
+      "p90": 1.0,
+      "min": 0.0,
+      "max": 1.0
+    },
+    "cache_read_share": {
+      "count": 216,
+      "mean": 0.6788762318171216,
+      "median": 0.7882742441214121,
+      "p10": 0.15606025798620213,
+      "p90": 0.9641169584856126,
+      "min": 0.0,
+      "max": 0.9893885889087582
+    },
+    "cache_write_share": {
+      "count": 216,
+      "mean": 0.15942191032151487,
+      "median": 0.043386068762250346,
+      "p10": 0.0,
+      "p90": 0.5204273102542047,
+      "min": 0.0,
+      "max": 0.999839897534422
+    },
+    "prompt_reuse_rate": {
+      "count": 216,
+      "mean": 0.6869508630393124,
+      "median": 0.7947498412123791,
+      "p10": 0.15876630503361927,
+      "p90": 0.971281214123751,
+      "min": 0.0,
+      "max": 0.9914821057091003
+    }
+  },
+  "top_primary_providers": [
+    [
+      "omni",
+      263
+    ],
+    [
+      "anthropic",
+      110
+    ],
+    [
+      "pi-claude-cli",
+      32
+    ],
+    [
+      "opencode",
+      27
+    ],
+    [
+      "openai",
+      10
+    ],
+    [
+      "nanogpt",
+      9
+    ],
+    [
+      "github-copilot",
+      3
+    ]
+  ],
+  "top_primary_models": [
+    [
+      "claude-opus-4-7",
+      100
+    ],
+    [
+      "claude-opus-4-8",
+      77
+    ],
+    [
+      "claude-opus-4-6",
+      46
+    ],
+    [
+      "opencode-zen/glm-5.1",
+      41
+    ],
+    [
+      "glm-5.1",
+      27
+    ],
+    [
+      "gpt-5.5-2026-04-23",
+      24
+    ],
+    [
+      "claude-sonnet-4-6",
+      21
+    ],
+    [
+      "gpt-5.4-2026-03-05",
+      10
+    ],
+    [
+      "anthropic/claude-opus-4-7",
+      10
+    ],
+    [
+      "gpt-5.5",
+      9
+    ]
+  ]
+}
\ No newline at end of file