docs: add #100, #110, #111, #139, #142, #143, #144 to unreleased changelog

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: jrob5756

CHANGELOG.md

@@ -16,6 +16,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   per-call inputs to sub-workflows evaluated against the parent context.
   When omitted, the parent's `workflow.input.*` is forwarded as before
   ([#109](https://github.com/microsoft/conductor/pull/109)).
+- `type: workflow` agents are now allowed inside `for_each` groups, enabling
+  dynamic fan-out to sub-workflows with per-iteration `input_mapping`. Each
+  iteration emits its own `subworkflow_started` / `subworkflow_completed`
+  events ([#110](https://github.com/microsoft/conductor/pull/110)).
+- Self-referential sub-workflows are now allowed; depth is bounded by the
+  global `MAX_SUBWORKFLOW_DEPTH` plus an optional per-agent `max_depth`
+  field on `AgentDef` ([#111](https://github.com/microsoft/conductor/pull/111)).
+
+### Changed
+- The dashboard's "context window remaining" bar now sources
+  `context_window_max` from each provider's SDK at runtime instead of a
+  hand-maintained static table. Values now reflect the actual cap the SDK
+  enforces (e.g. `claude-opus-4.6` reports 200K rather than the theoretical
+  1M; `gpt-5.x` reports 128K rather than 400K). The `context_window` field
+  on `ModelPricing` has been removed; pricing data continues to be
+  hand-maintained for cost calculation only
+  ([#144](https://github.com/microsoft/conductor/pull/144)).
 
 ### Fixed
 - Pass `streaming=True` to the Copilot SDK's `create_session` to prevent
@@ -24,6 +41,27 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   arguments (e.g., `create` with multi-KB `file_text`), the CLI executes
   the partial tool call, and the agent loops on the broken call until
   the wall-clock session limit fires ([#129](https://github.com/microsoft/conductor/pull/129)).
+- Build the Copilot prompt schema recursively from nested `output:`
+  definitions instead of flattening to top-level fields only. Nested object
+  properties, required keys, and array item schemas are now included in the
+  prompt-facing schema used for initial guidance and parse recovery
+  ([#100](https://github.com/microsoft/conductor/pull/100)).
+- Coerce Python literal `"True"` / `"False"` / `"None"` strings produced by
+  Jinja's default `str(bool)` rendering into native Python types when
+  building workflow output. Previously, `output: { matched: "{{ a == b }}" }`
+  produced the string `"False"` (truthy), causing downstream `when:`
+  comparisons against `false` to silently misbehave
+  ([#139](https://github.com/microsoft/conductor/pull/139)).
+- Pricing fuzzy match no longer silently inherits values across model
+  families. Names sharing a textual prefix with a known key (e.g.
+  `claude-opus-4.7` previously matched `claude-opus-4`) now require a `-`
+  delimiter; non-matching names return `None` and the dashboard hides the
+  cost field. A one-time warning is emitted per requested name on any
+  non-exact match ([#143](https://github.com/microsoft/conductor/pull/143)).
+- Run `uv tool update-shell` after `uv tool install` in both `install.ps1`
+  and `install.sh` so `conductor` is available on PATH in new shells, CI
+  agents, and IDE extensions after a fresh install
+  ([#142](https://github.com/microsoft/conductor/pull/142)).
 
 ## [0.1.10](https://github.com/microsoft/conductor/compare/v0.1.9...v0.1.10) - 2026-04-30